id,category,subcategory,difficulty,question,solution,tags
arb_0001,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_0002,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the umbrella from the bedroom and deliver it to the lab. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the umbrella (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the lab.
6. Deliver / place the umbrella in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0003,logical_reasoning,deduction,medium,No Rileys are in the library. Avery is in the library. Is Avery a Riley? Explain.,"No. The premise states that no Rileys are in the library. Since Avery is in the library, Avery cannot be a Riley.","deduction,syllogism,logic"
arb_0004,logical_reasoning,deduction,medium,"Either Jamie or Cameron took the pen, but not both. Jamie did not take the pen. Who took it?",Cameron took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0005,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0006,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0007,mathematical_reasoning,arithmetic,easy,Start with 32. Subtract 25. Then add 9. What is the final value? Show all steps.,"Step 1: 32 - 25 = 7
Step 2: 7 + 9 = 16
Final answer: 16","math,arithmetic,step-by-step"
arb_0008,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0009,logical_reasoning,deduction,easy,No Taylors are in the bedroom. Avery is in the bedroom. Is Avery a Taylor? Explain.,"No. The premise states that no Taylors are in the bedroom. Since Avery is in the bedroom, Avery cannot be a Taylor.","deduction,syllogism,logic"
arb_0010,mathematical_reasoning,arithmetic,easy,Start with 24. Add 30. Then multiply the result by 7. What is the final value? Show all steps.,"Step 1: 24 + 30 = 54
Step 2: 54 * 7 = 378
Final answer: 378","math,arithmetic,step-by-step"
arb_0011,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0012,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the server room. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0013,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0014,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0015,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0016,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0017,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0018,mathematical_reasoning,arithmetic,easy,Start with 27. Subtract 25. Then add 15. What is the final value? Show all steps.,"Step 1: 27 - 25 = 2
Step 2: 2 + 15 = 17
Final answer: 17","math,arithmetic,step-by-step"
arb_0019,tool_use,selection,easy,"Available tools: code_interpreter, email_client, calendar, database_query.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0020,logical_reasoning,deduction,medium,"Either Morgan or Casey took the umbrella, but not both. Morgan did not take the umbrella. Who took it?",Casey took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0021,commonsense_reasoning,everyday,easy,A blue cat is observed in the kitchen. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0022,commonsense_reasoning,everyday,easy,A white dog is observed in the library. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0023,mathematical_reasoning,arithmetic,easy,Compute (14 * 19) + 45. Explain the order of operations.,"Multiplication first: 14 * 19 = 266
Then addition: 266 + 45 = 311
Final answer: 311","math,arithmetic,step-by-step"
arb_0024,commonsense_reasoning,everyday,easy,A white owl is observed in the kitchen. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0025,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0026,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the laptop from the warehouse and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the laptop (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0027,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0028,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0029,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0030,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the pen from the lobby and deliver it to the bedroom. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the pen (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the bedroom.
6. Deliver / place the pen in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0031,logical_reasoning,deduction,medium,"If someone owns a charger, then they have access to the garage. Riley owns a charger. What can we conclude about Riley?","We can conclude that Riley has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_0032,mathematical_reasoning,arithmetic,easy,Compute (48 * 9) + 34. Explain the order of operations.,"Multiplication first: 48 * 9 = 432
Then addition: 432 + 34 = 466
Final answer: 466","math,arithmetic,step-by-step"
arb_0033,commonsense_reasoning,everyday,easy,A green owl is observed in the kitchen. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0034,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the umbrella from the lobby and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the umbrella (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the meeting room.
6. Deliver / place the umbrella in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0035,tool_use,selection,easy,"Available tools: calendar, web_search, file_system, code_interpreter, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0036,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the meeting room. Alex owns a umbrella. What can we conclude about Alex?","We can conclude that Alex has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_0037,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the pen from the library and deliver it to the meeting room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the pen (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the meeting room.
6. Deliver / place the pen in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0038,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the badge from the office and deliver it to the garage. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the badge (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the garage.
6. Deliver / place the badge in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0039,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0040,tool_use,selection,medium,"Available tools: terminal, code_interpreter, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0041,tool_use,selection,medium,"Available tools: web_search, browser, terminal, file_system.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0042,mathematical_reasoning,arithmetic,easy,Start with 37. Subtract 41. Then add 46. What is the final value? Show all steps.,"Step 1: 37 - 41 = -4
Step 2: -4 + 46 = 42
Final answer: 42","math,arithmetic,step-by-step"
arb_0043,tool_use,selection,easy,"Available tools: code_interpreter, file_system, email_client, calendar.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0044,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0045,tool_use,selection,easy,"Available tools: file_system, database_query, calendar, web_search, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0046,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the charger from the bedroom and deliver it to the lobby. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lobby.
6. Deliver / place the charger in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0047,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0048,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0049,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the wallet from the warehouse and deliver it to the lobby. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lobby.
6. Deliver / place the wallet in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0050,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0051,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the error logs.","causal,diagnosis,agentic"
arb_0052,tool_use,selection,easy,"Available tools: web_search, file_system, email_client, database_query, code_interpreter.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0053,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0054,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the phone from the lab and deliver it to the lobby. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the phone (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the lobby.
6. Deliver / place the phone in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0055,logical_reasoning,deduction,easy,All Alexs own a umbrella. Jamie is a Alex. Does Jamie own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Alexs own a umbrella' and the particular 'Jamie is a Alex', it follows by universal instantiation that Jamie owns a umbrella.","deduction,syllogism,logic"
arb_0056,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0057,tool_use,selection,easy,"Available tools: email_client, file_system, code_interpreter, terminal, calendar.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0058,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0059,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0060,logical_reasoning,deduction,easy,"If someone owns a key, then they have access to the lobby. Jordan owns a key. What can we conclude about Jordan?","We can conclude that Jordan has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_0061,tool_use,selection,medium,"Available tools: web_search, file_system, code_interpreter, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0062,mathematical_reasoning,arithmetic,easy,Start with 46. Subtract 4. Then add 29. What is the final value? Show all steps.,"Step 1: 46 - 4 = 42
Step 2: 42 + 29 = 71
Final answer: 71","math,arithmetic,step-by-step"
arb_0063,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0064,tool_use,selection,medium,"Available tools: code_interpreter, file_system, email_client.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0065,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0066,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0067,commonsense_reasoning,everyday,easy,A green wolf is observed in the server room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0068,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the umbrella from the kitchen and deliver it to the office. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the umbrella (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the office.
6. Deliver / place the umbrella in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0069,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0070,logical_reasoning,deduction,medium,"If someone owns a notebook, then they have access to the warehouse. Morgan owns a notebook. What can we conclude about Morgan?","We can conclude that Morgan has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_0071,tool_use,selection,easy,"Available tools: web_search, file_system, email_client, browser, code_interpreter.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0072,mathematical_reasoning,arithmetic,easy,Start with 20. Subtract 30. Then add 10. What is the final value? Show all steps.,"Step 1: 20 - 30 = -10
Step 2: -10 + 10 = 0
Final answer: 0","math,arithmetic,step-by-step"
arb_0073,tool_use,selection,easy,"Available tools: browser, calendar, code_interpreter, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0074,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0075,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0076,mathematical_reasoning,arithmetic,easy,Start with 17. Add 23. Then multiply the result by 38. What is the final value? Show all steps.,"Step 1: 17 + 23 = 40
Step 2: 40 * 38 = 1520
Final answer: 1520","math,arithmetic,step-by-step"
arb_0077,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0078,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0079,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the phone from the kitchen and deliver it to the bedroom. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the phone (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the bedroom.
6. Deliver / place the phone in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0080,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0081,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0082,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0083,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0084,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the pen from the meeting room and deliver it to the bedroom. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the pen (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the bedroom.
6. Deliver / place the pen in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0085,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0086,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0087,tool_use,selection,easy,"Available tools: code_interpreter, database_query, file_system.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0088,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0089,mathematical_reasoning,arithmetic,easy,Start with 17. Add 22. Then multiply the result by 23. What is the final value? Show all steps.,"Step 1: 17 + 22 = 39
Step 2: 39 * 23 = 897
Final answer: 897","math,arithmetic,step-by-step"
arb_0090,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0091,logical_reasoning,deduction,easy,All Morgans own a laptop. Taylor is a Morgan. Does Taylor own a laptop? Explain step by step.,"Yes. From the universal statement 'All Morgans own a laptop' and the particular 'Taylor is a Morgan', it follows by universal instantiation that Taylor owns a laptop.","deduction,syllogism,logic"
arb_0092,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0093,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0094,tool_use,selection,easy,"Available tools: email_client, web_search, database_query, calendar, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0095,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0096,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0097,tool_use,selection,medium,"Available tools: email_client, browser, database_query.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0098,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0099,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0100,mathematical_reasoning,arithmetic,easy,Compute (31 * 22) + 19. Explain the order of operations.,"Multiplication first: 31 * 22 = 682
Then addition: 682 + 19 = 701
Final answer: 701","math,arithmetic,step-by-step"
arb_0101,tool_use,selection,medium,"Available tools: code_interpreter, browser, calendar, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0102,commonsense_reasoning,everyday,easy,A yellow rabbit is observed in the lobby. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0103,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0104,commonsense_reasoning,everyday,easy,A green rabbit is observed in the warehouse. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0105,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the notebook from the meeting room and deliver it to the warehouse. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the notebook (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the warehouse.
6. Deliver / place the notebook in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0106,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0107,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the umbrella from the lab and deliver it to the meeting room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the umbrella (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the meeting room.
6. Deliver / place the umbrella in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0108,mathematical_reasoning,arithmetic,easy,Compute (2 * 14) + 31. Explain the order of operations.,"Multiplication first: 2 * 14 = 28
Then addition: 28 + 31 = 59
Final answer: 59","math,arithmetic,step-by-step"
arb_0109,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0110,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0111,tool_use,selection,medium,"Available tools: terminal, calendar, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0112,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the notebook from the garage and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the notebook (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the lab.
6. Deliver / place the notebook in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0113,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the key from the garage and deliver it to the bedroom. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the key (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the bedroom.
6. Deliver / place the key in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0114,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the garage. Casey owns a laptop. What can we conclude about Casey?","We can conclude that Casey has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_0115,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the umbrella from the lab and deliver it to the warehouse. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the warehouse.
6. Deliver / place the umbrella in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0116,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the notebook from the bedroom and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the notebook (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0117,mathematical_reasoning,arithmetic,easy,Compute (29 * 48) + 18. Explain the order of operations.,"Multiplication first: 29 * 48 = 1392
Then addition: 1392 + 18 = 1410
Final answer: 1410","math,arithmetic,step-by-step"
arb_0118,tool_use,selection,medium,"Available tools: browser, email_client, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0119,mathematical_reasoning,arithmetic,easy,Start with 49. Subtract 46. Then add 29. What is the final value? Show all steps.,"Step 1: 49 - 46 = 3
Step 2: 3 + 29 = 32
Final answer: 32","math,arithmetic,step-by-step"
arb_0120,tool_use,selection,easy,"Available tools: code_interpreter, database_query, browser.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0121,logical_reasoning,deduction,medium,"Either Avery or Alex took the umbrella, but not both. Avery did not take the umbrella. Who took it?",Alex took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0122,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0123,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0124,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0125,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the badge from the server room and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the meeting room.
6. Deliver / place the badge in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0126,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_0127,commonsense_reasoning,everyday,easy,A blue wolf is observed in the server room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0128,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0129,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0130,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0131,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0132,mathematical_reasoning,arithmetic,easy,Start with 49. Subtract 7. Then add 41. What is the final value? Show all steps.,"Step 1: 49 - 7 = 42
Step 2: 42 + 41 = 83
Final answer: 83","math,arithmetic,step-by-step"
arb_0133,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0134,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0135,logical_reasoning,deduction,easy,"Either Avery or Riley took the umbrella, but not both. Avery did not take the umbrella. Who took it?",Riley took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0136,commonsense_reasoning,everyday,easy,A purple bird is observed in the warehouse. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0137,mathematical_reasoning,arithmetic,easy,Start with 46. Add 23. Then multiply the result by 13. What is the final value? Show all steps.,"Step 1: 46 + 23 = 69
Step 2: 69 * 13 = 897
Final answer: 897","math,arithmetic,step-by-step"
arb_0138,logical_reasoning,deduction,easy,"Either Jamie or Morgan took the charger, but not both. Jamie did not take the charger. Who took it?",Morgan took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0139,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0140,logical_reasoning,deduction,easy,"If someone owns a pen, then they have access to the lab. Taylor owns a pen. What can we conclude about Taylor?","We can conclude that Taylor has access to the lab, by modus ponens.","deduction,syllogism,logic"
arb_0141,mathematical_reasoning,arithmetic,easy,Compute (21 * 12) + 2. Explain the order of operations.,"Multiplication first: 21 * 12 = 252
Then addition: 252 + 2 = 254
Final answer: 254","math,arithmetic,step-by-step"
arb_0142,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0143,logical_reasoning,deduction,medium,No Jamies are in the library. Casey is in the library. Is Casey a Jamie? Explain.,"No. The premise states that no Jamies are in the library. Since Casey is in the library, Casey cannot be a Jamie.","deduction,syllogism,logic"
arb_0144,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the umbrella from the kitchen and deliver it to the warehouse. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the umbrella (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the warehouse.
6. Deliver / place the umbrella in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0145,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the phone from the meeting room and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the phone (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the office.
6. Deliver / place the phone in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0146,mathematical_reasoning,arithmetic,easy,Start with 32. Subtract 7. Then add 27. What is the final value? Show all steps.,"Step 1: 32 - 7 = 25
Step 2: 25 + 27 = 52
Final answer: 52","math,arithmetic,step-by-step"
arb_0147,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0148,commonsense_reasoning,everyday,easy,A yellow owl is observed in the lab. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0149,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the laptop from the lab and deliver it to the kitchen. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the laptop (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the kitchen.
6. Deliver / place the laptop in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0150,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0151,tool_use,selection,medium,"Available tools: browser, code_interpreter, web_search, database_query, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0152,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the phone from the library and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0153,logical_reasoning,deduction,medium,"If someone owns a key, then they have access to the warehouse. Casey owns a key. What can we conclude about Casey?","We can conclude that Casey has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_0154,logical_reasoning,deduction,easy,"Either Avery or Riley took the book, but not both. Avery did not take the book. Who took it?",Riley took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0155,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0156,tool_use,selection,easy,"Available tools: browser, database_query, calendar.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0157,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0158,tool_use,selection,medium,"Available tools: web_search, code_interpreter, file_system.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0159,mathematical_reasoning,arithmetic,easy,Start with 28. Add 2. Then multiply the result by 8. What is the final value? Show all steps.,"Step 1: 28 + 2 = 30
Step 2: 30 * 8 = 240
Final answer: 240","math,arithmetic,step-by-step"
arb_0160,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0161,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the badge from the library and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the badge (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the server room.
6. Deliver / place the badge in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0162,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0163,mathematical_reasoning,arithmetic,easy,Start with 34. Subtract 28. Then add 10. What is the final value? Show all steps.,"Step 1: 34 - 28 = 6
Step 2: 6 + 10 = 16
Final answer: 16","math,arithmetic,step-by-step"
arb_0164,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the book from the library and deliver it to the lab. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the book (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the lab.
6. Deliver / place the book in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0165,logical_reasoning,deduction,easy,"If someone owns a notebook, then they have access to the meeting room. Cameron owns a notebook. What can we conclude about Cameron?","We can conclude that Cameron has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_0166,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0167,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the pen from the kitchen and deliver it to the garage. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the pen (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the garage.
6. Deliver / place the pen in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0168,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the charger from the meeting room and deliver it to the warehouse. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the charger (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the warehouse.
6. Deliver / place the charger in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0169,tool_use,selection,medium,"Available tools: file_system, web_search, database_query, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0170,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the laptop from the lobby and deliver it to the kitchen. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the laptop (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the kitchen.
6. Deliver / place the laptop in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0171,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0172,mathematical_reasoning,arithmetic,easy,Start with 34. Subtract 30. Then add 32. What is the final value? Show all steps.,"Step 1: 34 - 30 = 4
Step 2: 4 + 32 = 36
Final answer: 36","math,arithmetic,step-by-step"
arb_0173,tool_use,selection,medium,"Available tools: web_search, terminal, file_system, browser, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0174,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0175,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the pen from the lobby and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the pen (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0176,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0177,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the key from the bedroom and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the key (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0178,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_0179,logical_reasoning,deduction,medium,All Averys own a notebook. Morgan is a Avery. Does Morgan own a notebook? Explain step by step.,"Yes. From the universal statement 'All Averys own a notebook' and the particular 'Morgan is a Avery', it follows by universal instantiation that Morgan owns a notebook.","deduction,syllogism,logic"
arb_0180,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the laptop from the kitchen and deliver it to the server room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the server room.
6. Deliver / place the laptop in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0181,mathematical_reasoning,arithmetic,easy,Compute (49 * 9) + 39. Explain the order of operations.,"Multiplication first: 49 * 9 = 441
Then addition: 441 + 39 = 480
Final answer: 480","math,arithmetic,step-by-step"
arb_0182,logical_reasoning,deduction,easy,All Averys own a charger. Jamie is a Avery. Does Jamie own a charger? Explain step by step.,"Yes. From the universal statement 'All Averys own a charger' and the particular 'Jamie is a Avery', it follows by universal instantiation that Jamie owns a charger.","deduction,syllogism,logic"
arb_0183,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_0184,commonsense_reasoning,everyday,easy,A blue wolf is observed in the office. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0185,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the book from the bedroom and deliver it to the warehouse. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the book (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the warehouse.
6. Deliver / place the book in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0186,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the wallet from the meeting room and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the kitchen.
6. Deliver / place the wallet in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0187,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0188,tool_use,selection,medium,"Available tools: calendar, browser, web_search.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0189,commonsense_reasoning,everyday,easy,A green rabbit is observed in the bedroom. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0190,tool_use,selection,easy,"Available tools: code_interpreter, database_query, email_client, web_search, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0191,logical_reasoning,deduction,easy,No Caseys are in the kitchen. Quinn is in the kitchen. Is Quinn a Casey? Explain.,"No. The premise states that no Caseys are in the kitchen. Since Quinn is in the kitchen, Quinn cannot be a Casey.","deduction,syllogism,logic"
arb_0192,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0193,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the notebook from the lobby and deliver it to the lab. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the notebook (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the lab.
6. Deliver / place the notebook in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0194,mathematical_reasoning,arithmetic,easy,Start with 8. Add 14. Then multiply the result by 24. What is the final value? Show all steps.,"Step 1: 8 + 14 = 22
Step 2: 22 * 24 = 528
Final answer: 528","math,arithmetic,step-by-step"
arb_0195,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_0196,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0197,tool_use,selection,easy,"Available tools: email_client, code_interpreter, web_search, database_query, calendar.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0198,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0199,mathematical_reasoning,arithmetic,easy,Start with 24. Add 9. Then multiply the result by 36. What is the final value? Show all steps.,"Step 1: 24 + 9 = 33
Step 2: 33 * 36 = 1188
Final answer: 1188","math,arithmetic,step-by-step"
arb_0200,tool_use,selection,easy,"Available tools: file_system, database_query, terminal, calendar.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0201,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0202,logical_reasoning,deduction,easy,"If someone owns a badge, then they have access to the lobby. Riley owns a badge. What can we conclude about Riley?","We can conclude that Riley has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_0203,tool_use,selection,medium,"Available tools: email_client, database_query, terminal, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0204,logical_reasoning,deduction,medium,No Quinns are in the lobby. Jamie is in the lobby. Is Jamie a Quinn? Explain.,"No. The premise states that no Quinns are in the lobby. Since Jamie is in the lobby, Jamie cannot be a Quinn.","deduction,syllogism,logic"
arb_0205,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0206,commonsense_reasoning,everyday,easy,A red fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0207,mathematical_reasoning,arithmetic,easy,Start with 4. Subtract 42. Then add 19. What is the final value? Show all steps.,"Step 1: 4 - 42 = -38
Step 2: -38 + 19 = -19
Final answer: -19","math,arithmetic,step-by-step"
arb_0208,logical_reasoning,deduction,easy,"Either Jordan or Jamie took the book, but not both. Jordan did not take the book. Who took it?",Jamie took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0209,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0210,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the warehouse. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0211,mathematical_reasoning,arithmetic,easy,Start with 39. Add 14. Then multiply the result by 1. What is the final value? Show all steps.,"Step 1: 39 + 14 = 53
Step 2: 53 * 1 = 53
Final answer: 53","math,arithmetic,step-by-step"
arb_0212,commonsense_reasoning,everyday,easy,A purple bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0213,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0214,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_0215,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the key from the warehouse and deliver it to the lobby. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lobby.
6. Deliver / place the key in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0216,tool_use,selection,medium,"Available tools: database_query, web_search, email_client.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0217,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the key from the library and deliver it to the kitchen. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the key (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the kitchen.
6. Deliver / place the key in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0218,mathematical_reasoning,arithmetic,easy,Start with 37. Add 5. Then multiply the result by 20. What is the final value? Show all steps.,"Step 1: 37 + 5 = 42
Step 2: 42 * 20 = 840
Final answer: 840","math,arithmetic,step-by-step"
arb_0219,mathematical_reasoning,arithmetic,easy,Compute (29 * 6) + 35. Explain the order of operations.,"Multiplication first: 29 * 6 = 174
Then addition: 174 + 35 = 209
Final answer: 209","math,arithmetic,step-by-step"
arb_0220,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0221,logical_reasoning,deduction,easy,All Jamies own a charger. Jordan is a Jamie. Does Jordan own a charger? Explain step by step.,"Yes. From the universal statement 'All Jamies own a charger' and the particular 'Jordan is a Jamie', it follows by universal instantiation that Jordan owns a charger.","deduction,syllogism,logic"
arb_0222,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the badge from the warehouse and deliver it to the kitchen. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the badge (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the kitchen.
6. Deliver / place the badge in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0223,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0224,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0225,commonsense_reasoning,everyday,easy,A green bird is observed in the server room. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0226,tool_use,selection,easy,"Available tools: terminal, code_interpreter, web_search, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0227,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_0228,mathematical_reasoning,arithmetic,easy,Compute (30 * 31) + 3. Explain the order of operations.,"Multiplication first: 30 * 31 = 930
Then addition: 930 + 3 = 933
Final answer: 933","math,arithmetic,step-by-step"
arb_0229,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0230,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the wallet from the garage and deliver it to the lobby. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lobby.
6. Deliver / place the wallet in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0231,mathematical_reasoning,arithmetic,easy,Start with 9. Subtract 28. Then add 27. What is the final value? Show all steps.,"Step 1: 9 - 28 = -19
Step 2: -19 + 27 = 8
Final answer: 8","math,arithmetic,step-by-step"
arb_0232,commonsense_reasoning,everyday,easy,A orange fox is observed in the library. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0233,tool_use,selection,easy,"Available tools: email_client, database_query, web_search, file_system, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0234,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the library. Casey owns a umbrella. What can we conclude about Casey?","We can conclude that Casey has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_0235,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_0236,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0237,mathematical_reasoning,arithmetic,easy,Compute (15 * 42) + 44. Explain the order of operations.,"Multiplication first: 15 * 42 = 630
Then addition: 630 + 44 = 674
Final answer: 674","math,arithmetic,step-by-step"
arb_0238,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_0239,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the laptop from the warehouse and deliver it to the lab. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the laptop (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0240,logical_reasoning,deduction,medium,No Morgans are in the warehouse. Cameron is in the warehouse. Is Cameron a Morgan? Explain.,"No. The premise states that no Morgans are in the warehouse. Since Cameron is in the warehouse, Cameron cannot be a Morgan.","deduction,syllogism,logic"
arb_0241,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0242,mathematical_reasoning,arithmetic,easy,Compute (42 * 9) + 20. Explain the order of operations.,"Multiplication first: 42 * 9 = 378
Then addition: 378 + 20 = 398
Final answer: 398","math,arithmetic,step-by-step"
arb_0243,logical_reasoning,deduction,medium,All Taylors own a pen. Jamie is a Taylor. Does Jamie own a pen? Explain step by step.,"Yes. From the universal statement 'All Taylors own a pen' and the particular 'Jamie is a Taylor', it follows by universal instantiation that Jamie owns a pen.","deduction,syllogism,logic"
arb_0244,tool_use,selection,easy,"Available tools: calendar, code_interpreter, email_client, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0245,logical_reasoning,deduction,easy,"If someone owns a notebook, then they have access to the server room. Quinn owns a notebook. What can we conclude about Quinn?","We can conclude that Quinn has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_0246,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0247,mathematical_reasoning,arithmetic,easy,Start with 43. Add 41. Then multiply the result by 29. What is the final value? Show all steps.,"Step 1: 43 + 41 = 84
Step 2: 84 * 29 = 2436
Final answer: 2436","math,arithmetic,step-by-step"
arb_0248,mathematical_reasoning,arithmetic,easy,Compute (27 * 44) + 42. Explain the order of operations.,"Multiplication first: 27 * 44 = 1188
Then addition: 1188 + 42 = 1230
Final answer: 1230","math,arithmetic,step-by-step"
arb_0249,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the bedroom. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0250,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0251,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0252,mathematical_reasoning,arithmetic,easy,Start with 44. Add 13. Then multiply the result by 27. What is the final value? Show all steps.,"Step 1: 44 + 13 = 57
Step 2: 57 * 27 = 1539
Final answer: 1539","math,arithmetic,step-by-step"
arb_0253,mathematical_reasoning,arithmetic,easy,Start with 36. Add 4. Then multiply the result by 22. What is the final value? Show all steps.,"Step 1: 36 + 4 = 40
Step 2: 40 * 22 = 880
Final answer: 880","math,arithmetic,step-by-step"
arb_0254,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0255,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0256,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0257,mathematical_reasoning,arithmetic,easy,Start with 40. Add 41. Then multiply the result by 15. What is the final value? Show all steps.,"Step 1: 40 + 41 = 81
Step 2: 81 * 15 = 1215
Final answer: 1215","math,arithmetic,step-by-step"
arb_0258,mathematical_reasoning,arithmetic,easy,Compute (1 * 17) + 33. Explain the order of operations.,"Multiplication first: 1 * 17 = 17
Then addition: 17 + 33 = 50
Final answer: 50","math,arithmetic,step-by-step"
arb_0259,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0260,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0261,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0262,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0263,mathematical_reasoning,arithmetic,easy,Start with 18. Subtract 4. Then add 21. What is the final value? Show all steps.,"Step 1: 18 - 4 = 14
Step 2: 14 + 21 = 35
Final answer: 35","math,arithmetic,step-by-step"
arb_0264,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0265,commonsense_reasoning,everyday,easy,A orange wolf is observed in the server room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0266,tool_use,selection,easy,"Available tools: calendar, file_system, code_interpreter, web_search.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0267,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_0268,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the book from the office and deliver it to the meeting room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the book (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the meeting room.
6. Deliver / place the book in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0269,mathematical_reasoning,arithmetic,easy,Start with 13. Subtract 41. Then add 20. What is the final value? Show all steps.,"Step 1: 13 - 41 = -28
Step 2: -28 + 20 = -8
Final answer: -8","math,arithmetic,step-by-step"
arb_0270,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0271,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the wallet from the bedroom and deliver it to the garage. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the wallet (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0272,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the book from the garage and deliver it to the bedroom. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the book (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the bedroom.
6. Deliver / place the book in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0273,tool_use,selection,easy,"Available tools: email_client, terminal, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0274,mathematical_reasoning,arithmetic,easy,Start with 39. Subtract 15. Then add 23. What is the final value? Show all steps.,"Step 1: 39 - 15 = 24
Step 2: 24 + 23 = 47
Final answer: 47","math,arithmetic,step-by-step"
arb_0275,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0276,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_0277,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the server room. Quinn owns a umbrella. What can we conclude about Quinn?","We can conclude that Quinn has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_0278,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the badge from the lab and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the badge (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the garage.
6. Deliver / place the badge in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0279,tool_use,selection,medium,"Available tools: code_interpreter, web_search, email_client, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0280,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_0281,tool_use,selection,medium,"Available tools: browser, code_interpreter, file_system, email_client.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0282,tool_use,selection,easy,"Available tools: terminal, code_interpreter, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0283,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the badge from the office and deliver it to the bedroom. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the badge (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the bedroom.
6. Deliver / place the badge in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0284,logical_reasoning,deduction,easy,"Either Riley or Alex took the badge, but not both. Riley did not take the badge. Who took it?",Alex took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0285,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_0286,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0287,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0288,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0289,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0290,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0291,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0292,mathematical_reasoning,arithmetic,easy,Compute (23 * 27) + 39. Explain the order of operations.,"Multiplication first: 23 * 27 = 621
Then addition: 621 + 39 = 660
Final answer: 660","math,arithmetic,step-by-step"
arb_0293,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_0294,logical_reasoning,deduction,easy,No Jordans are in the kitchen. Quinn is in the kitchen. Is Quinn a Jordan? Explain.,"No. The premise states that no Jordans are in the kitchen. Since Quinn is in the kitchen, Quinn cannot be a Jordan.","deduction,syllogism,logic"
arb_0295,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the wallet from the kitchen and deliver it to the warehouse. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the wallet (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the warehouse.
6. Deliver / place the wallet in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0296,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the pen from the office and deliver it to the garage. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the garage.
6. Deliver / place the pen in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0297,tool_use,selection,medium,"Available tools: browser, file_system, email_client.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0298,mathematical_reasoning,arithmetic,easy,Compute (30 * 32) + 8. Explain the order of operations.,"Multiplication first: 30 * 32 = 960
Then addition: 960 + 8 = 968
Final answer: 968","math,arithmetic,step-by-step"
arb_0299,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the badge from the bedroom and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the badge (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the server room.
6. Deliver / place the badge in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0300,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0301,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_0302,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_0303,tool_use,selection,easy,"Available tools: file_system, browser, terminal, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0304,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the garage. Quinn owns a laptop. What can we conclude about Quinn?","We can conclude that Quinn has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_0305,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0306,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the wallet from the kitchen and deliver it to the bedroom. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the wallet (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the bedroom.
6. Deliver / place the wallet in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0307,tool_use,selection,medium,"Available tools: terminal, code_interpreter, email_client, database_query.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0308,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_0309,commonsense_reasoning,everyday,easy,A yellow bird is observed in the office. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a office (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0310,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0311,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0312,logical_reasoning,deduction,easy,No Morgans are in the warehouse. Riley is in the warehouse. Is Riley a Morgan? Explain.,"No. The premise states that no Morgans are in the warehouse. Since Riley is in the warehouse, Riley cannot be a Morgan.","deduction,syllogism,logic"
arb_0313,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_0314,mathematical_reasoning,arithmetic,easy,Start with 22. Add 9. Then multiply the result by 8. What is the final value? Show all steps.,"Step 1: 22 + 9 = 31
Step 2: 31 * 8 = 248
Final answer: 248","math,arithmetic,step-by-step"
arb_0315,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0316,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0317,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0318,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0319,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the umbrella from the bedroom and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the office.
6. Deliver / place the umbrella in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0320,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0321,mathematical_reasoning,arithmetic,easy,Start with 44. Subtract 45. Then add 36. What is the final value? Show all steps.,"Step 1: 44 - 45 = -1
Step 2: -1 + 36 = 35
Final answer: 35","math,arithmetic,step-by-step"
arb_0322,tool_use,selection,easy,"Available tools: code_interpreter, calendar, database_query.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0323,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the wallet from the lobby and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the wallet (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the server room.
6. Deliver / place the wallet in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0324,logical_reasoning,deduction,easy,No Jordans are in the library. Casey is in the library. Is Casey a Jordan? Explain.,"No. The premise states that no Jordans are in the library. Since Casey is in the library, Casey cannot be a Jordan.","deduction,syllogism,logic"
arb_0325,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0326,logical_reasoning,deduction,easy,"Either Quinn or Jordan took the notebook, but not both. Quinn did not take the notebook. Who took it?",Jordan took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0327,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_0328,mathematical_reasoning,arithmetic,easy,Start with 42. Subtract 12. Then add 41. What is the final value? Show all steps.,"Step 1: 42 - 12 = 30
Step 2: 30 + 41 = 71
Final answer: 71","math,arithmetic,step-by-step"
arb_0329,tool_use,selection,easy,"Available tools: database_query, file_system, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0330,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0331,mathematical_reasoning,arithmetic,easy,Start with 4. Subtract 36. Then add 19. What is the final value? Show all steps.,"Step 1: 4 - 36 = -32
Step 2: -32 + 19 = -13
Final answer: -13","math,arithmetic,step-by-step"
arb_0332,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the book from the office and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the book (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the server room.
6. Deliver / place the book in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0333,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0334,mathematical_reasoning,arithmetic,easy,Start with 21. Subtract 10. Then add 27. What is the final value? Show all steps.,"Step 1: 21 - 10 = 11
Step 2: 11 + 27 = 38
Final answer: 38","math,arithmetic,step-by-step"
arb_0335,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0336,tool_use,selection,medium,"Available tools: calendar, browser, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0337,logical_reasoning,deduction,medium,"Either Riley or Quinn took the umbrella, but not both. Riley did not take the umbrella. Who took it?",Quinn took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0338,logical_reasoning,deduction,medium,"If someone owns a badge, then they have access to the library. Avery owns a badge. What can we conclude about Avery?","We can conclude that Avery has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_0339,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0340,mathematical_reasoning,arithmetic,easy,Compute (33 * 2) + 8. Explain the order of operations.,"Multiplication first: 33 * 2 = 66
Then addition: 66 + 8 = 74
Final answer: 74","math,arithmetic,step-by-step"
arb_0341,commonsense_reasoning,everyday,easy,A red wolf is observed in the server room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0342,logical_reasoning,deduction,medium,"Either Avery or Casey took the book, but not both. Avery did not take the book. Who took it?",Casey took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0343,tool_use,selection,easy,"Available tools: calendar, web_search, terminal, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0344,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0345,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0346,commonsense_reasoning,everyday,easy,A black bird is observed in the server room. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0347,mathematical_reasoning,arithmetic,easy,Compute (31 * 28) + 2. Explain the order of operations.,"Multiplication first: 31 * 28 = 868
Then addition: 868 + 2 = 870
Final answer: 870","math,arithmetic,step-by-step"
arb_0348,mathematical_reasoning,arithmetic,easy,Start with 20. Subtract 37. Then add 29. What is the final value? Show all steps.,"Step 1: 20 - 37 = -17
Step 2: -17 + 29 = 12
Final answer: 12","math,arithmetic,step-by-step"
arb_0349,tool_use,selection,medium,"Available tools: web_search, email_client, terminal.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0350,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the laptop from the server room and deliver it to the office. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the laptop (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the office.
6. Deliver / place the laptop in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0351,tool_use,selection,medium,"Available tools: web_search, code_interpreter, email_client, calendar.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0352,logical_reasoning,deduction,easy,No Jordans are in the meeting room. Alex is in the meeting room. Is Alex a Jordan? Explain.,"No. The premise states that no Jordans are in the meeting room. Since Alex is in the meeting room, Alex cannot be a Jordan.","deduction,syllogism,logic"
arb_0353,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0354,tool_use,selection,medium,"Available tools: web_search, file_system, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0355,mathematical_reasoning,arithmetic,easy,Start with 33. Add 28. Then multiply the result by 27. What is the final value? Show all steps.,"Step 1: 33 + 28 = 61
Step 2: 61 * 27 = 1647
Final answer: 1647","math,arithmetic,step-by-step"
arb_0356,mathematical_reasoning,arithmetic,easy,Start with 42. Subtract 44. Then add 2. What is the final value? Show all steps.,"Step 1: 42 - 44 = -2
Step 2: -2 + 2 = 0
Final answer: 0","math,arithmetic,step-by-step"
arb_0357,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0358,tool_use,selection,medium,"Available tools: email_client, file_system, terminal, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0359,mathematical_reasoning,arithmetic,easy,Compute (2 * 49) + 18. Explain the order of operations.,"Multiplication first: 2 * 49 = 98
Then addition: 98 + 18 = 116
Final answer: 116","math,arithmetic,step-by-step"
arb_0360,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0361,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the laptop from the office and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the laptop (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the library.
6. Deliver / place the laptop in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0362,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0363,logical_reasoning,deduction,medium,All Caseys own a wallet. Alex is a Casey. Does Alex own a wallet? Explain step by step.,"Yes. From the universal statement 'All Caseys own a wallet' and the particular 'Alex is a Casey', it follows by universal instantiation that Alex owns a wallet.","deduction,syllogism,logic"
arb_0364,tool_use,selection,medium,"Available tools: terminal, database_query, browser, file_system, email_client.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0365,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the notebook from the office and deliver it to the meeting room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the notebook (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0366,logical_reasoning,deduction,easy,All Camerons own a key. Quinn is a Cameron. Does Quinn own a key? Explain step by step.,"Yes. From the universal statement 'All Camerons own a key' and the particular 'Quinn is a Cameron', it follows by universal instantiation that Quinn owns a key.","deduction,syllogism,logic"
arb_0367,logical_reasoning,deduction,easy,"If someone owns a charger, then they have access to the server room. Jamie owns a charger. What can we conclude about Jamie?","We can conclude that Jamie has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_0368,logical_reasoning,deduction,medium,All Morgans own a umbrella. Avery is a Morgan. Does Avery own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Morgans own a umbrella' and the particular 'Avery is a Morgan', it follows by universal instantiation that Avery owns a umbrella.","deduction,syllogism,logic"
arb_0369,mathematical_reasoning,arithmetic,easy,Start with 19. Subtract 47. Then add 46. What is the final value? Show all steps.,"Step 1: 19 - 47 = -28
Step 2: -28 + 46 = 18
Final answer: 18","math,arithmetic,step-by-step"
arb_0370,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0371,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the library. Alex owns a umbrella. What can we conclude about Alex?","We can conclude that Alex has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_0372,logical_reasoning,deduction,easy,"Either Morgan or Jordan took the laptop, but not both. Morgan did not take the laptop. Who took it?",Jordan took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0373,logical_reasoning,deduction,medium,All Averys own a umbrella. Alex is a Avery. Does Alex own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Averys own a umbrella' and the particular 'Alex is a Avery', it follows by universal instantiation that Alex owns a umbrella.","deduction,syllogism,logic"
arb_0374,commonsense_reasoning,everyday,easy,A purple owl is observed in the lobby. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0375,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0376,commonsense_reasoning,everyday,easy,A green cat is observed in the kitchen. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0377,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the umbrella from the kitchen and deliver it to the garage. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the umbrella (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the garage.
6. Deliver / place the umbrella in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0378,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0379,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0380,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0381,commonsense_reasoning,everyday,easy,A black rabbit is observed in the lobby. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0382,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0383,mathematical_reasoning,arithmetic,easy,Compute (15 * 2) + 36. Explain the order of operations.,"Multiplication first: 15 * 2 = 30
Then addition: 30 + 36 = 66
Final answer: 66","math,arithmetic,step-by-step"
arb_0384,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0385,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0386,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0387,commonsense_reasoning,everyday,easy,A purple cat is observed in the lobby. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0388,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the pen from the kitchen and deliver it to the office. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the pen (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the office.
6. Deliver / place the pen in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0389,commonsense_reasoning,everyday,easy,A green fox is observed in the warehouse. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0390,logical_reasoning,deduction,medium,"Either Taylor or Avery took the badge, but not both. Taylor did not take the badge. Who took it?",Avery took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0391,logical_reasoning,deduction,easy,"Either Morgan or Jamie took the badge, but not both. Morgan did not take the badge. Who took it?",Jamie took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0392,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0393,tool_use,selection,medium,"Available tools: database_query, email_client, file_system, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0394,mathematical_reasoning,arithmetic,easy,Compute (1 * 22) + 3. Explain the order of operations.,"Multiplication first: 1 * 22 = 22
Then addition: 22 + 3 = 25
Final answer: 25","math,arithmetic,step-by-step"
arb_0395,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_0396,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the laptop from the warehouse and deliver it to the bedroom. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the laptop (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the bedroom.
6. Deliver / place the laptop in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0397,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0398,mathematical_reasoning,arithmetic,easy,Start with 11. Add 36. Then multiply the result by 24. What is the final value? Show all steps.,"Step 1: 11 + 36 = 47
Step 2: 47 * 24 = 1128
Final answer: 1128","math,arithmetic,step-by-step"
arb_0399,logical_reasoning,deduction,easy,No Camerons are in the bedroom. Jordan is in the bedroom. Is Jordan a Cameron? Explain.,"No. The premise states that no Camerons are in the bedroom. Since Jordan is in the bedroom, Jordan cannot be a Cameron.","deduction,syllogism,logic"
arb_0400,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_0401,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0402,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0403,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0404,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0405,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0406,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_0407,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0408,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the phone from the meeting room and deliver it to the lobby. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the phone (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the lobby.
6. Deliver / place the phone in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0409,tool_use,selection,easy,"Available tools: database_query, web_search, terminal, calendar, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0410,mathematical_reasoning,arithmetic,easy,Compute (33 * 7) + 11. Explain the order of operations.,"Multiplication first: 33 * 7 = 231
Then addition: 231 + 11 = 242
Final answer: 242","math,arithmetic,step-by-step"
arb_0411,mathematical_reasoning,arithmetic,easy,Start with 45. Add 28. Then multiply the result by 5. What is the final value? Show all steps.,"Step 1: 45 + 28 = 73
Step 2: 73 * 5 = 365
Final answer: 365","math,arithmetic,step-by-step"
arb_0412,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0413,tool_use,selection,medium,"Available tools: web_search, database_query, email_client, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0414,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0415,mathematical_reasoning,arithmetic,easy,Compute (28 * 19) + 17. Explain the order of operations.,"Multiplication first: 28 * 19 = 532
Then addition: 532 + 17 = 549
Final answer: 549","math,arithmetic,step-by-step"
arb_0416,mathematical_reasoning,arithmetic,easy,Start with 10. Subtract 42. Then add 16. What is the final value? Show all steps.,"Step 1: 10 - 42 = -32
Step 2: -32 + 16 = -16
Final answer: -16","math,arithmetic,step-by-step"
arb_0417,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0418,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_0419,commonsense_reasoning,everyday,easy,A black wolf is observed in the kitchen. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0420,tool_use,selection,medium,"Available tools: browser, email_client, terminal, web_search, database_query.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0421,logical_reasoning,deduction,medium,No Alexs are in the server room. Cameron is in the server room. Is Cameron a Alex? Explain.,"No. The premise states that no Alexs are in the server room. Since Cameron is in the server room, Cameron cannot be a Alex.","deduction,syllogism,logic"
arb_0422,logical_reasoning,deduction,medium,No Jamies are in the meeting room. Casey is in the meeting room. Is Casey a Jamie? Explain.,"No. The premise states that no Jamies are in the meeting room. Since Casey is in the meeting room, Casey cannot be a Jamie.","deduction,syllogism,logic"
arb_0423,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0424,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0425,tool_use,selection,easy,"Available tools: terminal, browser, file_system.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0426,commonsense_reasoning,everyday,easy,A blue bird is observed in the library. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a library (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0427,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_0428,tool_use,selection,easy,"Available tools: email_client, database_query, browser, calendar.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0429,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the phone from the lobby and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the phone (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the kitchen.
6. Deliver / place the phone in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0430,mathematical_reasoning,arithmetic,easy,Start with 47. Subtract 19. Then add 46. What is the final value? Show all steps.,"Step 1: 47 - 19 = 28
Step 2: 28 + 46 = 74
Final answer: 74","math,arithmetic,step-by-step"
arb_0431,logical_reasoning,deduction,medium,"If someone owns a umbrella, then they have access to the library. Cameron owns a umbrella. What can we conclude about Cameron?","We can conclude that Cameron has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_0432,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0433,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_0434,logical_reasoning,deduction,medium,No Jordans are in the garage. Morgan is in the garage. Is Morgan a Jordan? Explain.,"No. The premise states that no Jordans are in the garage. Since Morgan is in the garage, Morgan cannot be a Jordan.","deduction,syllogism,logic"
arb_0435,logical_reasoning,deduction,easy,All Jamies own a book. Jordan is a Jamie. Does Jordan own a book? Explain step by step.,"Yes. From the universal statement 'All Jamies own a book' and the particular 'Jordan is a Jamie', it follows by universal instantiation that Jordan owns a book.","deduction,syllogism,logic"
arb_0436,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0437,commonsense_reasoning,everyday,easy,A black fox is observed in the library. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0438,tool_use,selection,medium,"Available tools: file_system, web_search, terminal, code_interpreter.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0439,mathematical_reasoning,arithmetic,easy,Compute (33 * 22) + 30. Explain the order of operations.,"Multiplication first: 33 * 22 = 726
Then addition: 726 + 30 = 756
Final answer: 756","math,arithmetic,step-by-step"
arb_0440,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0441,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0442,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0443,commonsense_reasoning,everyday,easy,A white dog is observed in the meeting room. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0444,mathematical_reasoning,arithmetic,easy,Compute (27 * 29) + 6. Explain the order of operations.,"Multiplication first: 27 * 29 = 783
Then addition: 783 + 6 = 789
Final answer: 789","math,arithmetic,step-by-step"
arb_0445,mathematical_reasoning,arithmetic,easy,Compute (20 * 10) + 9. Explain the order of operations.,"Multiplication first: 20 * 10 = 200
Then addition: 200 + 9 = 209
Final answer: 209","math,arithmetic,step-by-step"
arb_0446,logical_reasoning,deduction,easy,"If someone owns a wallet, then they have access to the lobby. Casey owns a wallet. What can we conclude about Casey?","We can conclude that Casey has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_0447,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0448,tool_use,selection,easy,"Available tools: calendar, file_system, email_client, browser, web_search.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0449,logical_reasoning,deduction,easy,"Either Jordan or Jamie took the book, but not both. Jordan did not take the book. Who took it?",Jamie took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0450,mathematical_reasoning,arithmetic,easy,Start with 19. Subtract 10. Then add 12. What is the final value? Show all steps.,"Step 1: 19 - 10 = 9
Step 2: 9 + 12 = 21
Final answer: 21","math,arithmetic,step-by-step"
arb_0451,logical_reasoning,deduction,medium,"If someone owns a umbrella, then they have access to the kitchen. Alex owns a umbrella. What can we conclude about Alex?","We can conclude that Alex has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_0452,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the laptop from the kitchen and deliver it to the lab. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0453,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0454,tool_use,selection,easy,"Available tools: browser, code_interpreter, database_query, calendar, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0455,mathematical_reasoning,arithmetic,easy,Start with 46. Add 28. Then multiply the result by 7. What is the final value? Show all steps.,"Step 1: 46 + 28 = 74
Step 2: 74 * 7 = 518
Final answer: 518","math,arithmetic,step-by-step"
arb_0456,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the pen from the warehouse and deliver it to the office. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the pen (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the office.
6. Deliver / place the pen in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0457,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0458,commonsense_reasoning,everyday,easy,A red owl is observed in the meeting room. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0459,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0460,logical_reasoning,deduction,easy,All Alexs own a charger. Jordan is a Alex. Does Jordan own a charger? Explain step by step.,"Yes. From the universal statement 'All Alexs own a charger' and the particular 'Jordan is a Alex', it follows by universal instantiation that Jordan owns a charger.","deduction,syllogism,logic"
arb_0461,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0462,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the key from the meeting room and deliver it to the warehouse. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the key (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the warehouse.
6. Deliver / place the key in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0463,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0464,commonsense_reasoning,everyday,easy,A orange dog is observed in the kitchen. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0465,mathematical_reasoning,arithmetic,easy,Start with 35. Add 29. Then multiply the result by 21. What is the final value? Show all steps.,"Step 1: 35 + 29 = 64
Step 2: 64 * 21 = 1344
Final answer: 1344","math,arithmetic,step-by-step"
arb_0466,logical_reasoning,deduction,medium,All Alexs own a laptop. Morgan is a Alex. Does Morgan own a laptop? Explain step by step.,"Yes. From the universal statement 'All Alexs own a laptop' and the particular 'Morgan is a Alex', it follows by universal instantiation that Morgan owns a laptop.","deduction,syllogism,logic"
arb_0467,commonsense_reasoning,everyday,easy,A red owl is observed in the warehouse. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0468,logical_reasoning,deduction,medium,All Quinns own a book. Alex is a Quinn. Does Alex own a book? Explain step by step.,"Yes. From the universal statement 'All Quinns own a book' and the particular 'Alex is a Quinn', it follows by universal instantiation that Alex owns a book.","deduction,syllogism,logic"
arb_0469,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the laptop from the office and deliver it to the server room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the laptop (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the server room.
6. Deliver / place the laptop in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0470,mathematical_reasoning,arithmetic,easy,Compute (35 * 11) + 16. Explain the order of operations.,"Multiplication first: 35 * 11 = 385
Then addition: 385 + 16 = 401
Final answer: 401","math,arithmetic,step-by-step"
arb_0471,tool_use,selection,medium,"Available tools: web_search, calendar, terminal, email_client.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0472,mathematical_reasoning,arithmetic,easy,Start with 4. Add 20. Then multiply the result by 19. What is the final value? Show all steps.,"Step 1: 4 + 20 = 24
Step 2: 24 * 19 = 456
Final answer: 456","math,arithmetic,step-by-step"
arb_0473,tool_use,selection,easy,"Available tools: browser, file_system, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0474,commonsense_reasoning,everyday,easy,A green fox is observed in the office. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0475,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0476,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the phone from the library and deliver it to the meeting room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the phone (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0477,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the book from the bedroom and deliver it to the lab. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the book (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the lab.
6. Deliver / place the book in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0478,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0479,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0480,commonsense_reasoning,everyday,easy,A red fish is observed in the garage. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0481,logical_reasoning,deduction,easy,"If someone owns a key, then they have access to the lab. Avery owns a key. What can we conclude about Avery?","We can conclude that Avery has access to the lab, by modus ponens.","deduction,syllogism,logic"
arb_0482,mathematical_reasoning,arithmetic,easy,Start with 4. Subtract 2. Then add 23. What is the final value? Show all steps.,"Step 1: 4 - 2 = 2
Step 2: 2 + 23 = 25
Final answer: 25","math,arithmetic,step-by-step"
arb_0483,logical_reasoning,deduction,easy,No Alexs are in the server room. Avery is in the server room. Is Avery a Alex? Explain.,"No. The premise states that no Alexs are in the server room. Since Avery is in the server room, Avery cannot be a Alex.","deduction,syllogism,logic"
arb_0484,logical_reasoning,deduction,medium,All Alexs own a charger. Jamie is a Alex. Does Jamie own a charger? Explain step by step.,"Yes. From the universal statement 'All Alexs own a charger' and the particular 'Jamie is a Alex', it follows by universal instantiation that Jamie owns a charger.","deduction,syllogism,logic"
arb_0485,logical_reasoning,deduction,easy,"If someone owns a wallet, then they have access to the meeting room. Cameron owns a wallet. What can we conclude about Cameron?","We can conclude that Cameron has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_0486,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the charger from the kitchen and deliver it to the garage. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the charger (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0487,logical_reasoning,deduction,medium,"If someone owns a phone, then they have access to the warehouse. Avery owns a phone. What can we conclude about Avery?","We can conclude that Avery has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_0488,tool_use,selection,easy,"Available tools: calendar, browser, terminal.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0489,tool_use,selection,medium,"Available tools: terminal, web_search, browser, file_system, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0490,mathematical_reasoning,arithmetic,easy,Start with 38. Subtract 16. Then add 13. What is the final value? Show all steps.,"Step 1: 38 - 16 = 22
Step 2: 22 + 13 = 35
Final answer: 35","math,arithmetic,step-by-step"
arb_0491,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the wallet from the server room and deliver it to the lobby. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the wallet (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lobby.
6. Deliver / place the wallet in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0492,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the pen from the meeting room and deliver it to the office. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the pen (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the office.
6. Deliver / place the pen in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0493,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_0494,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_0495,commonsense_reasoning,everyday,easy,A white fish is observed in the meeting room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0496,logical_reasoning,deduction,medium,"Either Avery or Quinn took the laptop, but not both. Avery did not take the laptop. Who took it?",Quinn took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0497,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0498,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_0499,tool_use,selection,medium,"Available tools: file_system, web_search, terminal, calendar, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0500,mathematical_reasoning,arithmetic,easy,Compute (37 * 44) + 39. Explain the order of operations.,"Multiplication first: 37 * 44 = 1628
Then addition: 1628 + 39 = 1667
Final answer: 1667","math,arithmetic,step-by-step"
arb_0501,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0502,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0503,commonsense_reasoning,everyday,easy,A green dog is observed in the meeting room. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0504,tool_use,selection,easy,"Available tools: terminal, code_interpreter, web_search, email_client.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0505,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0506,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0507,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0508,mathematical_reasoning,arithmetic,easy,Start with 28. Add 8. Then multiply the result by 40. What is the final value? Show all steps.,"Step 1: 28 + 8 = 36
Step 2: 36 * 40 = 1440
Final answer: 1440","math,arithmetic,step-by-step"
arb_0509,tool_use,selection,medium,"Available tools: code_interpreter, web_search, database_query, email_client.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0510,tool_use,selection,medium,"Available tools: email_client, calendar, code_interpreter, terminal, web_search.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0511,tool_use,selection,easy,"Available tools: web_search, calendar, code_interpreter.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0512,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0513,tool_use,selection,medium,"Available tools: email_client, calendar, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0514,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_0515,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_0516,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the key from the kitchen and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the key (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the server room.
6. Deliver / place the key in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0517,mathematical_reasoning,arithmetic,easy,Compute (6 * 24) + 11. Explain the order of operations.,"Multiplication first: 6 * 24 = 144
Then addition: 144 + 11 = 155
Final answer: 155","math,arithmetic,step-by-step"
arb_0518,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the book from the garage and deliver it to the meeting room. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the book (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the meeting room.
6. Deliver / place the book in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0519,commonsense_reasoning,everyday,easy,A red fish is observed in the garage. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0520,tool_use,selection,medium,"Available tools: code_interpreter, email_client, calendar.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0521,commonsense_reasoning,everyday,easy,A orange fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0522,logical_reasoning,deduction,medium,No Rileys are in the meeting room. Jordan is in the meeting room. Is Jordan a Riley? Explain.,"No. The premise states that no Rileys are in the meeting room. Since Jordan is in the meeting room, Jordan cannot be a Riley.","deduction,syllogism,logic"
arb_0523,commonsense_reasoning,everyday,easy,A green owl is observed in the kitchen. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0524,tool_use,selection,medium,"Available tools: file_system, email_client, calendar, browser, terminal.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0525,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0526,commonsense_reasoning,everyday,easy,A purple owl is observed in the library. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a library (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0527,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the phone from the library and deliver it to the garage. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the garage.
6. Deliver / place the phone in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0528,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the laptop from the office and deliver it to the meeting room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the laptop (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the meeting room.
6. Deliver / place the laptop in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0529,tool_use,selection,easy,"Available tools: code_interpreter, terminal, web_search.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0530,commonsense_reasoning,everyday,easy,A purple fox is observed in the lobby. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0531,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_0532,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0533,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_0534,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0535,mathematical_reasoning,arithmetic,easy,Compute (12 * 13) + 11. Explain the order of operations.,"Multiplication first: 12 * 13 = 156
Then addition: 156 + 11 = 167
Final answer: 167","math,arithmetic,step-by-step"
arb_0536,logical_reasoning,deduction,easy,No Taylors are in the library. Jamie is in the library. Is Jamie a Taylor? Explain.,"No. The premise states that no Taylors are in the library. Since Jamie is in the library, Jamie cannot be a Taylor.","deduction,syllogism,logic"
arb_0537,logical_reasoning,deduction,easy,"Either Casey or Alex took the notebook, but not both. Casey did not take the notebook. Who took it?",Alex took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0538,mathematical_reasoning,arithmetic,easy,Start with 45. Add 17. Then multiply the result by 33. What is the final value? Show all steps.,"Step 1: 45 + 17 = 62
Step 2: 62 * 33 = 2046
Final answer: 2046","math,arithmetic,step-by-step"
arb_0539,commonsense_reasoning,everyday,easy,A orange dog is observed in the warehouse. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0540,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the error logs.","causal,diagnosis,agentic"
arb_0541,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_0542,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the wallet from the warehouse and deliver it to the garage. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0543,logical_reasoning,deduction,easy,No Averys are in the server room. Riley is in the server room. Is Riley a Avery? Explain.,"No. The premise states that no Averys are in the server room. Since Riley is in the server room, Riley cannot be a Avery.","deduction,syllogism,logic"
arb_0544,mathematical_reasoning,arithmetic,easy,Start with 37. Add 33. Then multiply the result by 32. What is the final value? Show all steps.,"Step 1: 37 + 33 = 70
Step 2: 70 * 32 = 2240
Final answer: 2240","math,arithmetic,step-by-step"
arb_0545,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the umbrella from the office and deliver it to the warehouse. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the umbrella (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the warehouse.
6. Deliver / place the umbrella in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0546,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0547,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0548,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0549,logical_reasoning,deduction,easy,"Either Riley or Taylor took the book, but not both. Riley did not take the book. Who took it?",Taylor took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0550,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0551,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the laptop from the warehouse and deliver it to the meeting room. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the laptop (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the meeting room.
6. Deliver / place the laptop in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0552,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the badge from the office and deliver it to the lab. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the badge (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the lab.
6. Deliver / place the badge in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0553,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the charger from the library and deliver it to the lab. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the charger (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0554,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0555,mathematical_reasoning,arithmetic,easy,Compute (31 * 23) + 39. Explain the order of operations.,"Multiplication first: 31 * 23 = 713
Then addition: 713 + 39 = 752
Final answer: 752","math,arithmetic,step-by-step"
arb_0556,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0557,tool_use,selection,easy,"Available tools: terminal, code_interpreter, file_system, calendar, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0558,logical_reasoning,deduction,easy,All Caseys own a charger. Jamie is a Casey. Does Jamie own a charger? Explain step by step.,"Yes. From the universal statement 'All Caseys own a charger' and the particular 'Jamie is a Casey', it follows by universal instantiation that Jamie owns a charger.","deduction,syllogism,logic"
arb_0559,mathematical_reasoning,arithmetic,easy,Compute (36 * 18) + 10. Explain the order of operations.,"Multiplication first: 36 * 18 = 648
Then addition: 648 + 10 = 658
Final answer: 658","math,arithmetic,step-by-step"
arb_0560,tool_use,selection,medium,"Available tools: email_client, file_system, code_interpreter, terminal, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0561,tool_use,selection,easy,"Available tools: calendar, email_client, database_query, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0562,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0563,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_0564,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0565,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_0566,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0567,logical_reasoning,deduction,medium,"Either Alex or Casey took the umbrella, but not both. Alex did not take the umbrella. Who took it?",Casey took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0568,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0569,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0570,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0571,mathematical_reasoning,arithmetic,easy,Start with 10. Subtract 45. Then add 26. What is the final value? Show all steps.,"Step 1: 10 - 45 = -35
Step 2: -35 + 26 = -9
Final answer: -9","math,arithmetic,step-by-step"
arb_0572,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the laptop from the lobby and deliver it to the bedroom. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the bedroom.
6. Deliver / place the laptop in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0573,tool_use,selection,easy,"Available tools: file_system, calendar, browser.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0574,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the laptop from the bedroom and deliver it to the server room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the laptop (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the server room.
6. Deliver / place the laptop in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0575,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0576,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the notebook from the lobby and deliver it to the kitchen. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the notebook (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the kitchen.
6. Deliver / place the notebook in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0577,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the notebook from the warehouse and deliver it to the meeting room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the notebook (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0578,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0579,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0580,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0581,commonsense_reasoning,everyday,easy,A yellow cat is observed in the bedroom. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0582,logical_reasoning,deduction,easy,"Either Alex or Avery took the book, but not both. Alex did not take the book. Who took it?",Avery took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0583,tool_use,selection,medium,"Available tools: file_system, browser, web_search, calendar, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0584,tool_use,selection,medium,"Available tools: file_system, calendar, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0585,tool_use,selection,medium,"Available tools: web_search, browser, terminal, database_query, code_interpreter.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0586,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the umbrella from the lab and deliver it to the library. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the umbrella (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the library.
6. Deliver / place the umbrella in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0587,logical_reasoning,deduction,medium,All Alexs own a charger. Avery is a Alex. Does Avery own a charger? Explain step by step.,"Yes. From the universal statement 'All Alexs own a charger' and the particular 'Avery is a Alex', it follows by universal instantiation that Avery owns a charger.","deduction,syllogism,logic"
arb_0588,commonsense_reasoning,everyday,easy,A green wolf is observed in the bedroom. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0589,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the phone from the server room and deliver it to the bedroom. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the bedroom.
6. Deliver / place the phone in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0590,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0591,commonsense_reasoning,everyday,easy,A purple owl is observed in the library. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a library (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0592,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the key from the meeting room and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the kitchen.
6. Deliver / place the key in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0593,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the phone from the bedroom and deliver it to the warehouse. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the warehouse.
6. Deliver / place the phone in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0594,mathematical_reasoning,arithmetic,easy,Start with 1. Subtract 30. Then add 40. What is the final value? Show all steps.,"Step 1: 1 - 30 = -29
Step 2: -29 + 40 = 11
Final answer: 11","math,arithmetic,step-by-step"
arb_0595,logical_reasoning,deduction,easy,"If someone owns a badge, then they have access to the warehouse. Riley owns a badge. What can we conclude about Riley?","We can conclude that Riley has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_0596,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_0597,commonsense_reasoning,everyday,easy,A white cat is observed in the library. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a library (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0598,commonsense_reasoning,everyday,easy,A purple wolf is observed in the meeting room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0599,commonsense_reasoning,everyday,easy,A white cat is observed in the office. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a office (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0600,logical_reasoning,deduction,easy,"If someone owns a badge, then they have access to the server room. Taylor owns a badge. What can we conclude about Taylor?","We can conclude that Taylor has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_0601,mathematical_reasoning,arithmetic,easy,Start with 26. Add 13. Then multiply the result by 22. What is the final value? Show all steps.,"Step 1: 26 + 13 = 39
Step 2: 39 * 22 = 858
Final answer: 858","math,arithmetic,step-by-step"
arb_0602,tool_use,selection,easy,"Available tools: database_query, calendar, code_interpreter, email_client.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0603,commonsense_reasoning,everyday,easy,A purple bird is observed in the lobby. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0604,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0605,mathematical_reasoning,arithmetic,easy,Start with 1. Add 34. Then multiply the result by 37. What is the final value? Show all steps.,"Step 1: 1 + 34 = 35
Step 2: 35 * 37 = 1295
Final answer: 1295","math,arithmetic,step-by-step"
arb_0606,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the phone from the meeting room and deliver it to the library. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the phone (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the library.
6. Deliver / place the phone in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0607,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the phone from the bedroom and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the phone (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the office.
6. Deliver / place the phone in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0608,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0609,logical_reasoning,deduction,medium,No Rileys are in the warehouse. Jamie is in the warehouse. Is Jamie a Riley? Explain.,"No. The premise states that no Rileys are in the warehouse. Since Jamie is in the warehouse, Jamie cannot be a Riley.","deduction,syllogism,logic"
arb_0610,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0611,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0612,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0613,logical_reasoning,deduction,medium,All Caseys own a umbrella. Morgan is a Casey. Does Morgan own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Caseys own a umbrella' and the particular 'Morgan is a Casey', it follows by universal instantiation that Morgan owns a umbrella.","deduction,syllogism,logic"
arb_0614,commonsense_reasoning,everyday,easy,A purple fish is observed in the library. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0615,logical_reasoning,deduction,easy,"Either Jordan or Alex took the notebook, but not both. Jordan did not take the notebook. Who took it?",Alex took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0616,mathematical_reasoning,arithmetic,easy,Compute (3 * 27) + 31. Explain the order of operations.,"Multiplication first: 3 * 27 = 81
Then addition: 81 + 31 = 112
Final answer: 112","math,arithmetic,step-by-step"
arb_0617,commonsense_reasoning,everyday,easy,A green cat is observed in the meeting room. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0618,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0619,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_0620,tool_use,selection,medium,"Available tools: browser, web_search, email_client, code_interpreter.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0621,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0622,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0623,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_0624,tool_use,selection,easy,"Available tools: web_search, code_interpreter, calendar, browser, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0625,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0626,commonsense_reasoning,everyday,easy,A green wolf is observed in the lab. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0627,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_0628,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the charger from the kitchen and deliver it to the garage. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the charger (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0629,tool_use,selection,easy,"Available tools: database_query, browser, email_client, calendar.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0630,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0631,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0632,logical_reasoning,deduction,easy,All Morgans own a laptop. Taylor is a Morgan. Does Taylor own a laptop? Explain step by step.,"Yes. From the universal statement 'All Morgans own a laptop' and the particular 'Taylor is a Morgan', it follows by universal instantiation that Taylor owns a laptop.","deduction,syllogism,logic"
arb_0633,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the badge from the server room and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the office.
6. Deliver / place the badge in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0634,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0635,mathematical_reasoning,arithmetic,easy,Compute (25 * 31) + 18. Explain the order of operations.,"Multiplication first: 25 * 31 = 775
Then addition: 775 + 18 = 793
Final answer: 793","math,arithmetic,step-by-step"
arb_0636,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the book from the library and deliver it to the bedroom. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the book (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the bedroom.
6. Deliver / place the book in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0637,commonsense_reasoning,everyday,easy,A purple dog is observed in the garage. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0638,mathematical_reasoning,arithmetic,easy,Start with 46. Add 4. Then multiply the result by 34. What is the final value? Show all steps.,"Step 1: 46 + 4 = 50
Step 2: 50 * 34 = 1700
Final answer: 1700","math,arithmetic,step-by-step"
arb_0639,mathematical_reasoning,arithmetic,easy,Start with 20. Add 44. Then multiply the result by 39. What is the final value? Show all steps.,"Step 1: 20 + 44 = 64
Step 2: 64 * 39 = 2496
Final answer: 2496","math,arithmetic,step-by-step"
arb_0640,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the charger from the garage and deliver it to the warehouse. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the charger (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the warehouse.
6. Deliver / place the charger in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0641,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_0642,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_0643,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_0644,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0645,commonsense_reasoning,everyday,easy,A white cat is observed in the garage. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0646,mathematical_reasoning,arithmetic,easy,Start with 42. Subtract 4. Then add 21. What is the final value? Show all steps.,"Step 1: 42 - 4 = 38
Step 2: 38 + 21 = 59
Final answer: 59","math,arithmetic,step-by-step"
arb_0647,logical_reasoning,deduction,medium,"Either Quinn or Morgan took the badge, but not both. Quinn did not take the badge. Who took it?",Morgan took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0648,commonsense_reasoning,everyday,easy,A green owl is observed in the lobby. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0649,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0650,tool_use,selection,easy,"Available tools: email_client, calendar, file_system, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0651,tool_use,selection,easy,"Available tools: web_search, terminal, database_query, file_system.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0652,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0653,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0654,commonsense_reasoning,everyday,easy,A yellow rabbit is observed in the library. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a library (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0655,logical_reasoning,deduction,easy,"Either Cameron or Avery took the book, but not both. Cameron did not take the book. Who took it?",Avery took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0656,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the pen from the office and deliver it to the lobby. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the lobby.
6. Deliver / place the pen in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0657,tool_use,selection,medium,"Available tools: file_system, email_client, database_query, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0658,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0659,mathematical_reasoning,arithmetic,easy,Compute (36 * 42) + 8. Explain the order of operations.,"Multiplication first: 36 * 42 = 1512
Then addition: 1512 + 8 = 1520
Final answer: 1520","math,arithmetic,step-by-step"
arb_0660,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0661,commonsense_reasoning,everyday,easy,A yellow cat is observed in the server room. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0662,commonsense_reasoning,everyday,easy,A purple cat is observed in the bedroom. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0663,logical_reasoning,deduction,medium,No Rileys are in the garage. Avery is in the garage. Is Avery a Riley? Explain.,"No. The premise states that no Rileys are in the garage. Since Avery is in the garage, Avery cannot be a Riley.","deduction,syllogism,logic"
arb_0664,commonsense_reasoning,everyday,easy,A orange bird is observed in the office. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a office (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0665,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the key from the lobby and deliver it to the warehouse. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the warehouse.
6. Deliver / place the key in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0666,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the pen from the bedroom and deliver it to the warehouse. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the pen (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the warehouse.
6. Deliver / place the pen in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0667,commonsense_reasoning,everyday,easy,A white rabbit is observed in the kitchen. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0668,tool_use,selection,easy,"Available tools: file_system, terminal, database_query, calendar.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0669,tool_use,selection,easy,"Available tools: web_search, terminal, calendar.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0670,tool_use,selection,medium,"Available tools: database_query, calendar, browser.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0671,tool_use,selection,medium,"Available tools: code_interpreter, web_search, database_query, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0672,logical_reasoning,deduction,easy,No Rileys are in the library. Quinn is in the library. Is Quinn a Riley? Explain.,"No. The premise states that no Rileys are in the library. Since Quinn is in the library, Quinn cannot be a Riley.","deduction,syllogism,logic"
arb_0673,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the phone from the bedroom and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the phone (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the garage.
6. Deliver / place the phone in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0674,tool_use,selection,easy,"Available tools: terminal, web_search, file_system, database_query.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0675,tool_use,selection,medium,"Available tools: code_interpreter, file_system, email_client, terminal, web_search.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0676,logical_reasoning,deduction,medium,"Either Avery or Morgan took the umbrella, but not both. Avery did not take the umbrella. Who took it?",Morgan took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0677,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the phone from the lobby and deliver it to the warehouse. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the phone (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the warehouse.
6. Deliver / place the phone in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0678,mathematical_reasoning,arithmetic,easy,Compute (26 * 44) + 14. Explain the order of operations.,"Multiplication first: 26 * 44 = 1144
Then addition: 1144 + 14 = 1158
Final answer: 1158","math,arithmetic,step-by-step"
arb_0679,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the key from the meeting room and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the office.
6. Deliver / place the key in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0680,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the book from the meeting room and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the book (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the server room.
6. Deliver / place the book in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0681,tool_use,selection,medium,"Available tools: file_system, browser, database_query, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0682,tool_use,selection,medium,"Available tools: file_system, terminal, browser, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0683,logical_reasoning,deduction,easy,"Either Quinn or Taylor took the phone, but not both. Quinn did not take the phone. Who took it?",Taylor took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0684,logical_reasoning,deduction,easy,All Jamies own a badge. Riley is a Jamie. Does Riley own a badge? Explain step by step.,"Yes. From the universal statement 'All Jamies own a badge' and the particular 'Riley is a Jamie', it follows by universal instantiation that Riley owns a badge.","deduction,syllogism,logic"
arb_0685,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the phone from the bedroom and deliver it to the kitchen. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the phone (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the kitchen.
6. Deliver / place the phone in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0686,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0687,mathematical_reasoning,arithmetic,easy,Start with 46. Add 43. Then multiply the result by 30. What is the final value? Show all steps.,"Step 1: 46 + 43 = 89
Step 2: 89 * 30 = 2670
Final answer: 2670","math,arithmetic,step-by-step"
arb_0688,logical_reasoning,deduction,medium,No Caseys are in the garage. Taylor is in the garage. Is Taylor a Casey? Explain.,"No. The premise states that no Caseys are in the garage. Since Taylor is in the garage, Taylor cannot be a Casey.","deduction,syllogism,logic"
arb_0689,tool_use,selection,medium,"Available tools: calendar, database_query, terminal, email_client.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0690,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0691,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0692,logical_reasoning,deduction,medium,All Caseys own a wallet. Alex is a Casey. Does Alex own a wallet? Explain step by step.,"Yes. From the universal statement 'All Caseys own a wallet' and the particular 'Alex is a Casey', it follows by universal instantiation that Alex owns a wallet.","deduction,syllogism,logic"
arb_0693,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0694,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_0695,mathematical_reasoning,arithmetic,easy,Start with 33. Subtract 14. Then add 10. What is the final value? Show all steps.,"Step 1: 33 - 14 = 19
Step 2: 19 + 10 = 29
Final answer: 29","math,arithmetic,step-by-step"
arb_0696,logical_reasoning,deduction,easy,No Jordans are in the lobby. Morgan is in the lobby. Is Morgan a Jordan? Explain.,"No. The premise states that no Jordans are in the lobby. Since Morgan is in the lobby, Morgan cannot be a Jordan.","deduction,syllogism,logic"
arb_0697,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the wallet from the garage and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the meeting room.
6. Deliver / place the wallet in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0698,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the badge from the meeting room and deliver it to the bedroom. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the badge (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the bedroom.
6. Deliver / place the badge in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0699,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0700,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0701,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0702,logical_reasoning,deduction,medium,All Rileys own a notebook. Casey is a Riley. Does Casey own a notebook? Explain step by step.,"Yes. From the universal statement 'All Rileys own a notebook' and the particular 'Casey is a Riley', it follows by universal instantiation that Casey owns a notebook.","deduction,syllogism,logic"
arb_0703,mathematical_reasoning,arithmetic,easy,Compute (46 * 20) + 21. Explain the order of operations.,"Multiplication first: 46 * 20 = 920
Then addition: 920 + 21 = 941
Final answer: 941","math,arithmetic,step-by-step"
arb_0704,mathematical_reasoning,arithmetic,easy,Start with 4. Add 20. Then multiply the result by 25. What is the final value? Show all steps.,"Step 1: 4 + 20 = 24
Step 2: 24 * 25 = 600
Final answer: 600","math,arithmetic,step-by-step"
arb_0705,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0706,tool_use,selection,medium,"Available tools: code_interpreter, terminal, email_client, file_system.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0707,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0708,logical_reasoning,deduction,medium,No Alexs are in the warehouse. Avery is in the warehouse. Is Avery a Alex? Explain.,"No. The premise states that no Alexs are in the warehouse. Since Avery is in the warehouse, Avery cannot be a Alex.","deduction,syllogism,logic"
arb_0709,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0710,tool_use,selection,medium,"Available tools: terminal, browser, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0711,commonsense_reasoning,everyday,easy,A green rabbit is observed in the server room. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0712,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0713,logical_reasoning,deduction,easy,No Alexs are in the meeting room. Jamie is in the meeting room. Is Jamie a Alex? Explain.,"No. The premise states that no Alexs are in the meeting room. Since Jamie is in the meeting room, Jamie cannot be a Alex.","deduction,syllogism,logic"
arb_0714,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0715,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the charger from the kitchen and deliver it to the meeting room. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the charger (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the meeting room.
6. Deliver / place the charger in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0716,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0717,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0718,tool_use,selection,easy,"Available tools: email_client, terminal, web_search, file_system, database_query.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0719,mathematical_reasoning,arithmetic,easy,Start with 6. Add 43. Then multiply the result by 31. What is the final value? Show all steps.,"Step 1: 6 + 43 = 49
Step 2: 49 * 31 = 1519
Final answer: 1519","math,arithmetic,step-by-step"
arb_0720,mathematical_reasoning,arithmetic,easy,Start with 6. Add 11. Then multiply the result by 46. What is the final value? Show all steps.,"Step 1: 6 + 11 = 17
Step 2: 17 * 46 = 782
Final answer: 782","math,arithmetic,step-by-step"
arb_0721,commonsense_reasoning,everyday,easy,A purple owl is observed in the server room. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0722,commonsense_reasoning,everyday,easy,A white bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0723,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the phone from the office and deliver it to the meeting room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the phone (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0724,mathematical_reasoning,arithmetic,easy,Start with 20. Subtract 24. Then add 19. What is the final value? Show all steps.,"Step 1: 20 - 24 = -4
Step 2: -4 + 19 = 15
Final answer: 15","math,arithmetic,step-by-step"
arb_0725,tool_use,selection,medium,"Available tools: file_system, calendar, code_interpreter, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0726,logical_reasoning,deduction,medium,"If someone owns a laptop, then they have access to the library. Jordan owns a laptop. What can we conclude about Jordan?","We can conclude that Jordan has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_0727,tool_use,selection,easy,"Available tools: web_search, database_query, file_system, code_interpreter, browser.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0728,mathematical_reasoning,arithmetic,easy,Start with 2. Add 29. Then multiply the result by 10. What is the final value? Show all steps.,"Step 1: 2 + 29 = 31
Step 2: 31 * 10 = 310
Final answer: 310","math,arithmetic,step-by-step"
arb_0729,logical_reasoning,deduction,medium,All Rileys own a book. Cameron is a Riley. Does Cameron own a book? Explain step by step.,"Yes. From the universal statement 'All Rileys own a book' and the particular 'Cameron is a Riley', it follows by universal instantiation that Cameron owns a book.","deduction,syllogism,logic"
arb_0730,tool_use,selection,medium,"Available tools: file_system, database_query, browser.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0731,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the notebook from the server room and deliver it to the lab. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the notebook (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the lab.
6. Deliver / place the notebook in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0732,commonsense_reasoning,everyday,easy,A red bird is observed in the library. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a library (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0733,tool_use,selection,easy,"Available tools: code_interpreter, database_query, browser, email_client.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0734,logical_reasoning,deduction,medium,No Taylors are in the warehouse. Riley is in the warehouse. Is Riley a Taylor? Explain.,"No. The premise states that no Taylors are in the warehouse. Since Riley is in the warehouse, Riley cannot be a Taylor.","deduction,syllogism,logic"
arb_0735,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_0736,logical_reasoning,deduction,medium,No Quinns are in the warehouse. Riley is in the warehouse. Is Riley a Quinn? Explain.,"No. The premise states that no Quinns are in the warehouse. Since Riley is in the warehouse, Riley cannot be a Quinn.","deduction,syllogism,logic"
arb_0737,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_0738,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0739,commonsense_reasoning,everyday,easy,A black fish is observed in the bedroom. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0740,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0741,commonsense_reasoning,everyday,easy,A orange owl is observed in the meeting room. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0742,commonsense_reasoning,everyday,easy,A purple dog is observed in the lobby. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0743,mathematical_reasoning,arithmetic,easy,Compute (30 * 13) + 17. Explain the order of operations.,"Multiplication first: 30 * 13 = 390
Then addition: 390 + 17 = 407
Final answer: 407","math,arithmetic,step-by-step"
arb_0744,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0745,commonsense_reasoning,everyday,easy,A white bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0746,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the charger from the warehouse and deliver it to the office. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the office.
6. Deliver / place the charger in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0747,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0748,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the notebook from the meeting room and deliver it to the office. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the notebook (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the office.
6. Deliver / place the notebook in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0749,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0750,commonsense_reasoning,everyday,easy,A blue wolf is observed in the warehouse. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0751,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0752,logical_reasoning,deduction,medium,"If someone owns a charger, then they have access to the bedroom. Alex owns a charger. What can we conclude about Alex?","We can conclude that Alex has access to the bedroom, by modus ponens.","deduction,syllogism,logic"
arb_0753,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the badge from the server room and deliver it to the kitchen. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the kitchen.
6. Deliver / place the badge in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0754,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the meeting room. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0755,commonsense_reasoning,everyday,easy,A green wolf is observed in the library. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0756,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the error logs.","causal,diagnosis,agentic"
arb_0757,tool_use,selection,easy,"Available tools: calendar, database_query, terminal, browser, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0758,tool_use,selection,easy,"Available tools: file_system, browser, web_search, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0759,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_0760,tool_use,selection,medium,"Available tools: calendar, browser, terminal, email_client, file_system.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0761,commonsense_reasoning,everyday,easy,A blue dog is observed in the library. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0762,logical_reasoning,deduction,medium,"Either Alex or Cameron took the laptop, but not both. Alex did not take the laptop. Who took it?",Cameron took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0763,logical_reasoning,deduction,medium,No Rileys are in the warehouse. Cameron is in the warehouse. Is Cameron a Riley? Explain.,"No. The premise states that no Rileys are in the warehouse. Since Cameron is in the warehouse, Cameron cannot be a Riley.","deduction,syllogism,logic"
arb_0764,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the charger from the meeting room and deliver it to the kitchen. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the kitchen.
6. Deliver / place the charger in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0765,commonsense_reasoning,everyday,easy,A red fish is observed in the server room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0766,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the wallet from the kitchen and deliver it to the warehouse. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the wallet (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the warehouse.
6. Deliver / place the wallet in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0767,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0768,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0769,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0770,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0771,mathematical_reasoning,arithmetic,easy,Start with 7. Subtract 10. Then add 42. What is the final value? Show all steps.,"Step 1: 7 - 10 = -3
Step 2: -3 + 42 = 39
Final answer: 39","math,arithmetic,step-by-step"
arb_0772,commonsense_reasoning,everyday,easy,A orange bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0773,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the badge from the server room and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the kitchen.
6. Deliver / place the badge in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0774,logical_reasoning,deduction,medium,"Either Morgan or Alex took the wallet, but not both. Morgan did not take the wallet. Who took it?",Alex took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0775,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the book from the server room and deliver it to the lobby. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the book (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the lobby.
6. Deliver / place the book in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0776,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0777,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_0778,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the umbrella from the server room and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the umbrella (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the meeting room.
6. Deliver / place the umbrella in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0779,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the umbrella from the office and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the umbrella (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the library.
6. Deliver / place the umbrella in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0780,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the umbrella from the bedroom and deliver it to the office. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the umbrella (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the office.
6. Deliver / place the umbrella in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0781,tool_use,selection,medium,"Available tools: database_query, terminal, browser, web_search, calendar.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0782,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0783,logical_reasoning,deduction,easy,"If someone owns a pen, then they have access to the kitchen. Quinn owns a pen. What can we conclude about Quinn?","We can conclude that Quinn has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_0784,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the phone from the lobby and deliver it to the office. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the phone (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the office.
6. Deliver / place the phone in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0785,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_0786,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0787,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the book from the office and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the book (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the library.
6. Deliver / place the book in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0788,tool_use,selection,medium,"Available tools: database_query, browser, web_search, email_client, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0789,logical_reasoning,deduction,medium,"Either Casey or Avery took the key, but not both. Casey did not take the key. Who took it?",Avery took the key. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0790,commonsense_reasoning,everyday,easy,A black bird is observed in the kitchen. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0791,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0792,mathematical_reasoning,arithmetic,easy,Start with 7. Subtract 31. Then add 16. What is the final value? Show all steps.,"Step 1: 7 - 31 = -24
Step 2: -24 + 16 = -8
Final answer: -8","math,arithmetic,step-by-step"
arb_0793,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0794,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0795,logical_reasoning,deduction,easy,"If someone owns a book, then they have access to the office. Cameron owns a book. What can we conclude about Cameron?","We can conclude that Cameron has access to the office, by modus ponens.","deduction,syllogism,logic"
arb_0796,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the pen from the warehouse and deliver it to the lab. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the pen (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the lab.
6. Deliver / place the pen in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0797,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the bedroom. Jamie owns a laptop. What can we conclude about Jamie?","We can conclude that Jamie has access to the bedroom, by modus ponens.","deduction,syllogism,logic"
arb_0798,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the charger from the warehouse and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the server room.
6. Deliver / place the charger in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0799,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_0800,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0801,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_0802,mathematical_reasoning,arithmetic,easy,Start with 5. Subtract 39. Then add 40. What is the final value? Show all steps.,"Step 1: 5 - 39 = -34
Step 2: -34 + 40 = 6
Final answer: 6","math,arithmetic,step-by-step"
arb_0803,logical_reasoning,deduction,medium,"Either Avery or Jordan took the badge, but not both. Avery did not take the badge. Who took it?",Jordan took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0804,logical_reasoning,deduction,easy,"Either Riley or Casey took the phone, but not both. Riley did not take the phone. Who took it?",Casey took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0805,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0806,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0807,logical_reasoning,deduction,easy,No Rileys are in the warehouse. Taylor is in the warehouse. Is Taylor a Riley? Explain.,"No. The premise states that no Rileys are in the warehouse. Since Taylor is in the warehouse, Taylor cannot be a Riley.","deduction,syllogism,logic"
arb_0808,logical_reasoning,deduction,medium,"Either Jamie or Morgan took the notebook, but not both. Jamie did not take the notebook. Who took it?",Morgan took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0809,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0810,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the badge from the garage and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the badge (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the library.
6. Deliver / place the badge in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0811,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0812,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0813,tool_use,selection,easy,"Available tools: browser, email_client, web_search, file_system, calendar.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0814,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0815,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0816,mathematical_reasoning,arithmetic,easy,Start with 33. Subtract 13. Then add 32. What is the final value? Show all steps.,"Step 1: 33 - 13 = 20
Step 2: 20 + 32 = 52
Final answer: 52","math,arithmetic,step-by-step"
arb_0817,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0818,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0819,mathematical_reasoning,arithmetic,easy,Start with 34. Subtract 10. Then add 8. What is the final value? Show all steps.,"Step 1: 34 - 10 = 24
Step 2: 24 + 8 = 32
Final answer: 32","math,arithmetic,step-by-step"
arb_0820,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0821,logical_reasoning,deduction,medium,"If someone owns a umbrella, then they have access to the lobby. Jamie owns a umbrella. What can we conclude about Jamie?","We can conclude that Jamie has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_0822,logical_reasoning,deduction,easy,"If someone owns a wallet, then they have access to the kitchen. Avery owns a wallet. What can we conclude about Avery?","We can conclude that Avery has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_0823,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_0824,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_0825,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0826,mathematical_reasoning,arithmetic,easy,Start with 30. Subtract 28. Then add 6. What is the final value? Show all steps.,"Step 1: 30 - 28 = 2
Step 2: 2 + 6 = 8
Final answer: 8","math,arithmetic,step-by-step"
arb_0827,tool_use,selection,medium,"Available tools: file_system, email_client, browser.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0828,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the notebook from the kitchen and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the notebook (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the lab.
6. Deliver / place the notebook in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0829,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0830,mathematical_reasoning,arithmetic,easy,Compute (16 * 48) + 38. Explain the order of operations.,"Multiplication first: 16 * 48 = 768
Then addition: 768 + 38 = 806
Final answer: 806","math,arithmetic,step-by-step"
arb_0831,mathematical_reasoning,arithmetic,easy,Start with 29. Add 27. Then multiply the result by 1. What is the final value? Show all steps.,"Step 1: 29 + 27 = 56
Step 2: 56 * 1 = 56
Final answer: 56","math,arithmetic,step-by-step"
arb_0832,tool_use,selection,easy,"Available tools: database_query, email_client, code_interpreter, terminal, web_search.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0833,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the charger from the lobby and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the charger (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0834,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_0835,mathematical_reasoning,arithmetic,easy,Start with 45. Add 2. Then multiply the result by 49. What is the final value? Show all steps.,"Step 1: 45 + 2 = 47
Step 2: 47 * 49 = 2303
Final answer: 2303","math,arithmetic,step-by-step"
arb_0836,logical_reasoning,deduction,easy,No Caseys are in the office. Jamie is in the office. Is Jamie a Casey? Explain.,"No. The premise states that no Caseys are in the office. Since Jamie is in the office, Jamie cannot be a Casey.","deduction,syllogism,logic"
arb_0837,logical_reasoning,deduction,medium,"Either Morgan or Alex took the phone, but not both. Morgan did not take the phone. Who took it?",Alex took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0838,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0839,commonsense_reasoning,everyday,easy,A green fox is observed in the lobby. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0840,logical_reasoning,deduction,medium,No Averys are in the office. Alex is in the office. Is Alex a Avery? Explain.,"No. The premise states that no Averys are in the office. Since Alex is in the office, Alex cannot be a Avery.","deduction,syllogism,logic"
arb_0841,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0842,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the wallet from the lobby and deliver it to the lab. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the wallet (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lab.
6. Deliver / place the wallet in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0843,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0844,tool_use,selection,easy,"Available tools: web_search, email_client, terminal.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0845,tool_use,selection,medium,"Available tools: code_interpreter, file_system, calendar, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0846,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0847,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0848,logical_reasoning,deduction,medium,"Either Taylor or Jamie took the charger, but not both. Taylor did not take the charger. Who took it?",Jamie took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0849,commonsense_reasoning,everyday,easy,A black wolf is observed in the lobby. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0850,mathematical_reasoning,arithmetic,easy,Compute (1 * 28) + 42. Explain the order of operations.,"Multiplication first: 1 * 28 = 28
Then addition: 28 + 42 = 70
Final answer: 70","math,arithmetic,step-by-step"
arb_0851,mathematical_reasoning,arithmetic,easy,Start with 21. Add 34. Then multiply the result by 38. What is the final value? Show all steps.,"Step 1: 21 + 34 = 55
Step 2: 55 * 38 = 2090
Final answer: 2090","math,arithmetic,step-by-step"
arb_0852,mathematical_reasoning,arithmetic,easy,Start with 15. Add 32. Then multiply the result by 24. What is the final value? Show all steps.,"Step 1: 15 + 32 = 47
Step 2: 47 * 24 = 1128
Final answer: 1128","math,arithmetic,step-by-step"
arb_0853,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0854,logical_reasoning,deduction,medium,"If someone owns a pen, then they have access to the kitchen. Avery owns a pen. What can we conclude about Avery?","We can conclude that Avery has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_0855,logical_reasoning,deduction,easy,All Jordans own a laptop. Morgan is a Jordan. Does Morgan own a laptop? Explain step by step.,"Yes. From the universal statement 'All Jordans own a laptop' and the particular 'Morgan is a Jordan', it follows by universal instantiation that Morgan owns a laptop.","deduction,syllogism,logic"
arb_0856,logical_reasoning,deduction,easy,All Rileys own a notebook. Quinn is a Riley. Does Quinn own a notebook? Explain step by step.,"Yes. From the universal statement 'All Rileys own a notebook' and the particular 'Quinn is a Riley', it follows by universal instantiation that Quinn owns a notebook.","deduction,syllogism,logic"
arb_0857,logical_reasoning,deduction,easy,"Either Taylor or Quinn took the book, but not both. Taylor did not take the book. Who took it?",Quinn took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0858,tool_use,selection,medium,"Available tools: database_query, web_search, calendar, terminal.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0859,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0860,commonsense_reasoning,everyday,easy,A orange owl is observed in the office. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a office (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0861,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0862,tool_use,selection,easy,"Available tools: database_query, terminal, web_search, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0863,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0864,logical_reasoning,deduction,easy,All Jordans own a laptop. Quinn is a Jordan. Does Quinn own a laptop? Explain step by step.,"Yes. From the universal statement 'All Jordans own a laptop' and the particular 'Quinn is a Jordan', it follows by universal instantiation that Quinn owns a laptop.","deduction,syllogism,logic"
arb_0865,mathematical_reasoning,arithmetic,easy,Start with 40. Add 44. Then multiply the result by 32. What is the final value? Show all steps.,"Step 1: 40 + 44 = 84
Step 2: 84 * 32 = 2688
Final answer: 2688","math,arithmetic,step-by-step"
arb_0866,tool_use,selection,medium,"Available tools: terminal, web_search, browser, code_interpreter, file_system.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0867,tool_use,selection,easy,"Available tools: code_interpreter, terminal, file_system, browser, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0868,tool_use,selection,medium,"Available tools: terminal, web_search, database_query.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0869,logical_reasoning,deduction,medium,All Alexs own a umbrella. Jamie is a Alex. Does Jamie own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Alexs own a umbrella' and the particular 'Jamie is a Alex', it follows by universal instantiation that Jamie owns a umbrella.","deduction,syllogism,logic"
arb_0870,commonsense_reasoning,everyday,easy,A orange rabbit is observed in the lab. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0871,mathematical_reasoning,arithmetic,easy,Start with 48. Subtract 5. Then add 9. What is the final value? Show all steps.,"Step 1: 48 - 5 = 43
Step 2: 43 + 9 = 52
Final answer: 52","math,arithmetic,step-by-step"
arb_0872,tool_use,selection,medium,"Available tools: email_client, code_interpreter, calendar, terminal, file_system.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0873,tool_use,selection,easy,"Available tools: email_client, web_search, browser, code_interpreter, database_query.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0874,tool_use,selection,medium,"Available tools: calendar, email_client, file_system, database_query, code_interpreter.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0875,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0876,mathematical_reasoning,arithmetic,easy,Start with 13. Subtract 11. Then add 22. What is the final value? Show all steps.,"Step 1: 13 - 11 = 2
Step 2: 2 + 22 = 24
Final answer: 24","math,arithmetic,step-by-step"
arb_0877,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the notebook from the library and deliver it to the warehouse. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the notebook (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the warehouse.
6. Deliver / place the notebook in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0878,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0879,tool_use,selection,medium,"Available tools: code_interpreter, file_system, database_query, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0880,mathematical_reasoning,arithmetic,easy,Start with 34. Add 22. Then multiply the result by 38. What is the final value? Show all steps.,"Step 1: 34 + 22 = 56
Step 2: 56 * 38 = 2128
Final answer: 2128","math,arithmetic,step-by-step"
arb_0881,mathematical_reasoning,arithmetic,easy,Compute (44 * 7) + 16. Explain the order of operations.,"Multiplication first: 44 * 7 = 308
Then addition: 308 + 16 = 324
Final answer: 324","math,arithmetic,step-by-step"
arb_0882,tool_use,selection,medium,"Available tools: terminal, file_system, email_client.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0883,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_0884,commonsense_reasoning,everyday,easy,A yellow cat is observed in the warehouse. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0885,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_0886,tool_use,selection,easy,"Available tools: database_query, calendar, code_interpreter.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0887,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_0888,mathematical_reasoning,arithmetic,easy,Compute (43 * 38) + 14. Explain the order of operations.,"Multiplication first: 43 * 38 = 1634
Then addition: 1634 + 14 = 1648
Final answer: 1648","math,arithmetic,step-by-step"
arb_0889,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0890,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the laptop from the bedroom and deliver it to the warehouse. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the laptop (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0891,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0892,mathematical_reasoning,arithmetic,easy,Compute (28 * 11) + 21. Explain the order of operations.,"Multiplication first: 28 * 11 = 308
Then addition: 308 + 21 = 329
Final answer: 329","math,arithmetic,step-by-step"
arb_0893,commonsense_reasoning,everyday,easy,A yellow bird is observed in the meeting room. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0894,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the book from the office and deliver it to the library. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the book (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the library.
6. Deliver / place the book in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0895,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_0896,mathematical_reasoning,arithmetic,easy,Start with 14. Subtract 30. Then add 39. What is the final value? Show all steps.,"Step 1: 14 - 30 = -16
Step 2: -16 + 39 = 23
Final answer: 23","math,arithmetic,step-by-step"
arb_0897,tool_use,selection,easy,"Available tools: file_system, database_query, calendar.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0898,tool_use,selection,easy,"Available tools: file_system, browser, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0899,commonsense_reasoning,everyday,easy,A purple bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0900,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0901,tool_use,selection,medium,"Available tools: database_query, web_search, calendar, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0902,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_0903,commonsense_reasoning,everyday,easy,A black cat is observed in the library. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a library (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0904,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the charger from the lobby and deliver it to the server room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the charger (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the server room.
6. Deliver / place the charger in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0905,tool_use,selection,medium,"Available tools: code_interpreter, database_query, calendar.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0906,logical_reasoning,deduction,medium,No Jordans are in the library. Cameron is in the library. Is Cameron a Jordan? Explain.,"No. The premise states that no Jordans are in the library. Since Cameron is in the library, Cameron cannot be a Jordan.","deduction,syllogism,logic"
arb_0907,tool_use,selection,easy,"Available tools: calendar, browser, email_client, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0908,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_0909,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the umbrella from the lab and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the umbrella (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the garage.
6. Deliver / place the umbrella in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0910,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_0911,logical_reasoning,deduction,medium,"Either Alex or Taylor took the key, but not both. Alex did not take the key. Who took it?",Taylor took the key. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0912,logical_reasoning,deduction,easy,"Either Cameron or Jamie took the umbrella, but not both. Cameron did not take the umbrella. Who took it?",Jamie took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0913,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the wallet from the library and deliver it to the server room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the server room.
6. Deliver / place the wallet in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0914,mathematical_reasoning,arithmetic,easy,Start with 12. Add 5. Then multiply the result by 38. What is the final value? Show all steps.,"Step 1: 12 + 5 = 17
Step 2: 17 * 38 = 646
Final answer: 646","math,arithmetic,step-by-step"
arb_0915,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0916,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the pen from the office and deliver it to the library. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the library.
6. Deliver / place the pen in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0917,tool_use,selection,easy,"Available tools: email_client, terminal, code_interpreter, calendar, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0918,logical_reasoning,deduction,medium,No Averys are in the lobby. Jamie is in the lobby. Is Jamie a Avery? Explain.,"No. The premise states that no Averys are in the lobby. Since Jamie is in the lobby, Jamie cannot be a Avery.","deduction,syllogism,logic"
arb_0919,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the laptop from the lobby and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the meeting room.
6. Deliver / place the laptop in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0920,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the bedroom. Riley owns a umbrella. What can we conclude about Riley?","We can conclude that Riley has access to the bedroom, by modus ponens.","deduction,syllogism,logic"
arb_0921,tool_use,selection,easy,"Available tools: terminal, browser, calendar, database_query, web_search.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0922,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the laptop from the server room and deliver it to the kitchen. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the laptop (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the kitchen.
6. Deliver / place the laptop in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0923,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the meeting room. Riley owns a laptop. What can we conclude about Riley?","We can conclude that Riley has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_0924,commonsense_reasoning,everyday,easy,A red cat is observed in the lab. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0925,commonsense_reasoning,everyday,easy,A red owl is observed in the server room. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0926,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0927,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0928,tool_use,selection,easy,"Available tools: database_query, terminal, browser, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0929,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the umbrella from the lobby and deliver it to the garage. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the umbrella (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the garage.
6. Deliver / place the umbrella in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0930,logical_reasoning,deduction,easy,"Either Avery or Morgan took the charger, but not both. Avery did not take the charger. Who took it?",Morgan took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0931,tool_use,selection,medium,"Available tools: code_interpreter, browser, file_system, web_search.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0932,tool_use,selection,medium,"Available tools: terminal, calendar, database_query, web_search.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0933,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0934,tool_use,selection,medium,"Available tools: browser, calendar, code_interpreter.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0935,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0936,tool_use,selection,medium,"Available tools: database_query, calendar, web_search, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0937,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the server room. Jamie owns a umbrella. What can we conclude about Jamie?","We can conclude that Jamie has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_0938,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the wallet from the lab and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the wallet (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the meeting room.
6. Deliver / place the wallet in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0939,tool_use,selection,medium,"Available tools: web_search, email_client, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0940,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the key from the warehouse and deliver it to the meeting room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0941,commonsense_reasoning,everyday,easy,A white rabbit is observed in the office. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a office (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0942,tool_use,selection,medium,"Available tools: file_system, database_query, email_client, browser.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0943,tool_use,selection,easy,"Available tools: terminal, file_system, browser, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0944,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0945,mathematical_reasoning,arithmetic,easy,Start with 37. Add 2. Then multiply the result by 48. What is the final value? Show all steps.,"Step 1: 37 + 2 = 39
Step 2: 39 * 48 = 1872
Final answer: 1872","math,arithmetic,step-by-step"
arb_0946,logical_reasoning,deduction,easy,All Quinns own a charger. Jamie is a Quinn. Does Jamie own a charger? Explain step by step.,"Yes. From the universal statement 'All Quinns own a charger' and the particular 'Jamie is a Quinn', it follows by universal instantiation that Jamie owns a charger.","deduction,syllogism,logic"
arb_0947,commonsense_reasoning,everyday,easy,A blue bird is observed in the lobby. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0948,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0949,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the laptop from the lobby and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the laptop (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the library.
6. Deliver / place the laptop in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0950,logical_reasoning,deduction,medium,"If someone owns a key, then they have access to the meeting room. Quinn owns a key. What can we conclude about Quinn?","We can conclude that Quinn has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_0951,commonsense_reasoning,everyday,easy,A green cat is observed in the lobby. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0952,commonsense_reasoning,everyday,easy,A purple dog is observed in the warehouse. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0953,tool_use,selection,easy,"Available tools: email_client, terminal, browser, calendar.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0954,mathematical_reasoning,arithmetic,easy,Start with 42. Add 30. Then multiply the result by 6. What is the final value? Show all steps.,"Step 1: 42 + 30 = 72
Step 2: 72 * 6 = 432
Final answer: 432","math,arithmetic,step-by-step"
arb_0955,commonsense_reasoning,everyday,easy,A white fish is observed in the kitchen. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0956,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the pen from the bedroom and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the pen (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the meeting room.
6. Deliver / place the pen in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0957,logical_reasoning,deduction,easy,No Camerons are in the meeting room. Casey is in the meeting room. Is Casey a Cameron? Explain.,"No. The premise states that no Camerons are in the meeting room. Since Casey is in the meeting room, Casey cannot be a Cameron.","deduction,syllogism,logic"
arb_0958,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0959,logical_reasoning,deduction,medium,All Quinns own a book. Cameron is a Quinn. Does Cameron own a book? Explain step by step.,"Yes. From the universal statement 'All Quinns own a book' and the particular 'Cameron is a Quinn', it follows by universal instantiation that Cameron owns a book.","deduction,syllogism,logic"
arb_0960,tool_use,selection,medium,"Available tools: database_query, calendar, web_search, file_system, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0961,tool_use,selection,easy,"Available tools: browser, file_system, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0962,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_0963,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_0964,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_0965,tool_use,selection,easy,"Available tools: file_system, terminal, database_query, email_client.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0966,tool_use,selection,easy,"Available tools: web_search, email_client, terminal.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0967,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_0968,mathematical_reasoning,arithmetic,easy,Compute (28 * 5) + 7. Explain the order of operations.,"Multiplication first: 28 * 5 = 140
Then addition: 140 + 7 = 147
Final answer: 147","math,arithmetic,step-by-step"
arb_0969,mathematical_reasoning,arithmetic,easy,Compute (42 * 14) + 35. Explain the order of operations.,"Multiplication first: 42 * 14 = 588
Then addition: 588 + 35 = 623
Final answer: 623","math,arithmetic,step-by-step"
arb_0970,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_0971,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0972,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0973,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the phone from the lobby and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the phone (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the office.
6. Deliver / place the phone in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0974,tool_use,selection,easy,"Available tools: terminal, browser, calendar.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0975,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the wallet from the library and deliver it to the bedroom. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the bedroom.
6. Deliver / place the wallet in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0976,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0977,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the wallet from the lab and deliver it to the meeting room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the wallet (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the meeting room.
6. Deliver / place the wallet in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0978,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0979,mathematical_reasoning,arithmetic,easy,Compute (32 * 33) + 38. Explain the order of operations.,"Multiplication first: 32 * 33 = 1056
Then addition: 1056 + 38 = 1094
Final answer: 1094","math,arithmetic,step-by-step"
arb_0980,logical_reasoning,deduction,easy,All Rileys own a key. Jordan is a Riley. Does Jordan own a key? Explain step by step.,"Yes. From the universal statement 'All Rileys own a key' and the particular 'Jordan is a Riley', it follows by universal instantiation that Jordan owns a key.","deduction,syllogism,logic"
arb_0981,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_0982,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the charger from the lobby and deliver it to the garage. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the charger (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0983,commonsense_reasoning,everyday,easy,A blue dog is observed in the server room. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0984,logical_reasoning,deduction,easy,"If someone owns a badge, then they have access to the meeting room. Taylor owns a badge. What can we conclude about Taylor?","We can conclude that Taylor has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_0985,logical_reasoning,deduction,medium,"Either Morgan or Jordan took the book, but not both. Morgan did not take the book. Who took it?",Jordan took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0986,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0987,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the key from the lab and deliver it to the warehouse. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the key (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the warehouse.
6. Deliver / place the key in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_0988,logical_reasoning,deduction,easy,All Quinns own a laptop. Jamie is a Quinn. Does Jamie own a laptop? Explain step by step.,"Yes. From the universal statement 'All Quinns own a laptop' and the particular 'Jamie is a Quinn', it follows by universal instantiation that Jamie owns a laptop.","deduction,syllogism,logic"
arb_0989,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_0990,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0991,tool_use,selection,medium,"Available tools: file_system, email_client, terminal, code_interpreter.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_0992,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_0993,commonsense_reasoning,everyday,easy,A green fish is observed in the kitchen. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_0994,logical_reasoning,deduction,easy,All Jordans own a wallet. Morgan is a Jordan. Does Morgan own a wallet? Explain step by step.,"Yes. From the universal statement 'All Jordans own a wallet' and the particular 'Morgan is a Jordan', it follows by universal instantiation that Morgan owns a wallet.","deduction,syllogism,logic"
arb_0995,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_0996,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_0997,logical_reasoning,deduction,easy,"Either Cameron or Jamie took the key, but not both. Cameron did not take the key. Who took it?",Jamie took the key. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_0998,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_0999,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the charger from the meeting room and deliver it to the bedroom. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the charger (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the bedroom.
6. Deliver / place the charger in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1000,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1001,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the wallet from the garage and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the wallet (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the office.
6. Deliver / place the wallet in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1002,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_1003,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the laptop from the kitchen and deliver it to the library. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the library.
6. Deliver / place the laptop in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1004,tool_use,selection,medium,"Available tools: file_system, code_interpreter, database_query, email_client, terminal.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1005,commonsense_reasoning,everyday,easy,A black fox is observed in the lab. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1006,commonsense_reasoning,everyday,easy,A purple fox is observed in the server room. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1007,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the badge from the meeting room and deliver it to the lobby. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the badge (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the lobby.
6. Deliver / place the badge in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1008,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1009,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_1010,logical_reasoning,deduction,medium,No Taylors are in the office. Alex is in the office. Is Alex a Taylor? Explain.,"No. The premise states that no Taylors are in the office. Since Alex is in the office, Alex cannot be a Taylor.","deduction,syllogism,logic"
arb_1011,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1012,mathematical_reasoning,arithmetic,easy,Compute (46 * 41) + 43. Explain the order of operations.,"Multiplication first: 46 * 41 = 1886
Then addition: 1886 + 43 = 1929
Final answer: 1929","math,arithmetic,step-by-step"
arb_1013,mathematical_reasoning,arithmetic,easy,Start with 8. Add 7. Then multiply the result by 35. What is the final value? Show all steps.,"Step 1: 8 + 7 = 15
Step 2: 15 * 35 = 525
Final answer: 525","math,arithmetic,step-by-step"
arb_1014,logical_reasoning,deduction,medium,"If someone owns a phone, then they have access to the library. Morgan owns a phone. What can we conclude about Morgan?","We can conclude that Morgan has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_1015,commonsense_reasoning,everyday,easy,A black fox is observed in the lab. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1016,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1017,mathematical_reasoning,arithmetic,easy,Compute (19 * 37) + 4. Explain the order of operations.,"Multiplication first: 19 * 37 = 703
Then addition: 703 + 4 = 707
Final answer: 707","math,arithmetic,step-by-step"
arb_1018,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the pen from the bedroom and deliver it to the warehouse. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the pen (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the warehouse.
6. Deliver / place the pen in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1019,mathematical_reasoning,arithmetic,easy,Start with 28. Add 36. Then multiply the result by 29. What is the final value? Show all steps.,"Step 1: 28 + 36 = 64
Step 2: 64 * 29 = 1856
Final answer: 1856","math,arithmetic,step-by-step"
arb_1020,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1021,logical_reasoning,deduction,medium,No Rileys are in the kitchen. Morgan is in the kitchen. Is Morgan a Riley? Explain.,"No. The premise states that no Rileys are in the kitchen. Since Morgan is in the kitchen, Morgan cannot be a Riley.","deduction,syllogism,logic"
arb_1022,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the charger from the garage and deliver it to the warehouse. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the charger (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the warehouse.
6. Deliver / place the charger in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1023,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the charger from the warehouse and deliver it to the garage. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1024,tool_use,selection,easy,"Available tools: calendar, database_query, web_search, browser.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1025,mathematical_reasoning,arithmetic,easy,Compute (8 * 1) + 46. Explain the order of operations.,"Multiplication first: 8 * 1 = 8
Then addition: 8 + 46 = 54
Final answer: 54","math,arithmetic,step-by-step"
arb_1026,mathematical_reasoning,arithmetic,easy,Start with 38. Subtract 36. Then add 34. What is the final value? Show all steps.,"Step 1: 38 - 36 = 2
Step 2: 2 + 34 = 36
Final answer: 36","math,arithmetic,step-by-step"
arb_1027,mathematical_reasoning,arithmetic,easy,Start with 15. Subtract 4. Then add 18. What is the final value? Show all steps.,"Step 1: 15 - 4 = 11
Step 2: 11 + 18 = 29
Final answer: 29","math,arithmetic,step-by-step"
arb_1028,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the key from the garage and deliver it to the bedroom. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the key (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the bedroom.
6. Deliver / place the key in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1029,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the key from the library and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the key (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lab.
6. Deliver / place the key in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1030,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1031,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1032,tool_use,selection,medium,"Available tools: email_client, file_system, terminal, database_query, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1033,logical_reasoning,deduction,easy,No Jordans are in the lab. Casey is in the lab. Is Casey a Jordan? Explain.,"No. The premise states that no Jordans are in the lab. Since Casey is in the lab, Casey cannot be a Jordan.","deduction,syllogism,logic"
arb_1034,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1035,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1036,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the book from the lobby and deliver it to the kitchen. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the book (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the kitchen.
6. Deliver / place the book in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1037,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1038,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1039,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_1040,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the wallet from the library and deliver it to the lobby. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lobby.
6. Deliver / place the wallet in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1041,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1042,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1043,mathematical_reasoning,arithmetic,easy,Start with 19. Subtract 3. Then add 32. What is the final value? Show all steps.,"Step 1: 19 - 3 = 16
Step 2: 16 + 32 = 48
Final answer: 48","math,arithmetic,step-by-step"
arb_1044,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the key from the kitchen and deliver it to the library. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the library.
6. Deliver / place the key in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1045,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1046,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the key from the garage and deliver it to the lobby. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lobby.
6. Deliver / place the key in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1047,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1048,mathematical_reasoning,arithmetic,easy,Start with 28. Add 39. Then multiply the result by 17. What is the final value? Show all steps.,"Step 1: 28 + 39 = 67
Step 2: 67 * 17 = 1139
Final answer: 1139","math,arithmetic,step-by-step"
arb_1049,commonsense_reasoning,everyday,easy,A purple dog is observed in the server room. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1050,logical_reasoning,deduction,easy,"If someone owns a charger, then they have access to the bedroom. Alex owns a charger. What can we conclude about Alex?","We can conclude that Alex has access to the bedroom, by modus ponens.","deduction,syllogism,logic"
arb_1051,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1052,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1053,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the charger from the server room and deliver it to the bedroom. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the charger (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the bedroom.
6. Deliver / place the charger in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1054,logical_reasoning,deduction,medium,"If someone owns a badge, then they have access to the office. Quinn owns a badge. What can we conclude about Quinn?","We can conclude that Quinn has access to the office, by modus ponens.","deduction,syllogism,logic"
arb_1055,tool_use,selection,medium,"Available tools: calendar, email_client, web_search, terminal, file_system.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1056,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1057,mathematical_reasoning,arithmetic,easy,Start with 30. Subtract 34. Then add 44. What is the final value? Show all steps.,"Step 1: 30 - 34 = -4
Step 2: -4 + 44 = 40
Final answer: 40","math,arithmetic,step-by-step"
arb_1058,commonsense_reasoning,everyday,easy,A yellow bird is observed in the warehouse. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1059,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1060,mathematical_reasoning,arithmetic,easy,Compute (10 * 32) + 34. Explain the order of operations.,"Multiplication first: 10 * 32 = 320
Then addition: 320 + 34 = 354
Final answer: 354","math,arithmetic,step-by-step"
arb_1061,mathematical_reasoning,arithmetic,easy,Compute (3 * 11) + 48. Explain the order of operations.,"Multiplication first: 3 * 11 = 33
Then addition: 33 + 48 = 81
Final answer: 81","math,arithmetic,step-by-step"
arb_1062,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1063,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1064,logical_reasoning,deduction,medium,No Averys are in the server room. Quinn is in the server room. Is Quinn a Avery? Explain.,"No. The premise states that no Averys are in the server room. Since Quinn is in the server room, Quinn cannot be a Avery.","deduction,syllogism,logic"
arb_1065,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the pen from the server room and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the pen (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the lab.
6. Deliver / place the pen in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1066,logical_reasoning,deduction,easy,No Jordans are in the office. Casey is in the office. Is Casey a Jordan? Explain.,"No. The premise states that no Jordans are in the office. Since Casey is in the office, Casey cannot be a Jordan.","deduction,syllogism,logic"
arb_1067,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the phone from the library and deliver it to the warehouse. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the phone (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the warehouse.
6. Deliver / place the phone in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1068,logical_reasoning,deduction,medium,No Jordans are in the lab. Taylor is in the lab. Is Taylor a Jordan? Explain.,"No. The premise states that no Jordans are in the lab. Since Taylor is in the lab, Taylor cannot be a Jordan.","deduction,syllogism,logic"
arb_1069,tool_use,selection,medium,"Available tools: file_system, browser, code_interpreter.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1070,logical_reasoning,deduction,medium,"Either Riley or Quinn took the badge, but not both. Riley did not take the badge. Who took it?",Quinn took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1071,mathematical_reasoning,arithmetic,easy,Start with 32. Subtract 18. Then add 48. What is the final value? Show all steps.,"Step 1: 32 - 18 = 14
Step 2: 14 + 48 = 62
Final answer: 62","math,arithmetic,step-by-step"
arb_1072,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_1073,logical_reasoning,deduction,easy,No Caseys are in the kitchen. Riley is in the kitchen. Is Riley a Casey? Explain.,"No. The premise states that no Caseys are in the kitchen. Since Riley is in the kitchen, Riley cannot be a Casey.","deduction,syllogism,logic"
arb_1074,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_1075,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1076,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1077,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1078,logical_reasoning,deduction,medium,No Jamies are in the library. Jordan is in the library. Is Jordan a Jamie? Explain.,"No. The premise states that no Jamies are in the library. Since Jordan is in the library, Jordan cannot be a Jamie.","deduction,syllogism,logic"
arb_1079,logical_reasoning,deduction,medium,All Taylors own a laptop. Jordan is a Taylor. Does Jordan own a laptop? Explain step by step.,"Yes. From the universal statement 'All Taylors own a laptop' and the particular 'Jordan is a Taylor', it follows by universal instantiation that Jordan owns a laptop.","deduction,syllogism,logic"
arb_1080,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the umbrella from the lab and deliver it to the library. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the umbrella (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the library.
6. Deliver / place the umbrella in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1081,commonsense_reasoning,everyday,easy,A yellow fish is observed in the lobby. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1082,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1083,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_1084,mathematical_reasoning,arithmetic,easy,Start with 41. Add 43. Then multiply the result by 16. What is the final value? Show all steps.,"Step 1: 41 + 43 = 84
Step 2: 84 * 16 = 1344
Final answer: 1344","math,arithmetic,step-by-step"
arb_1085,mathematical_reasoning,arithmetic,easy,Compute (4 * 48) + 5. Explain the order of operations.,"Multiplication first: 4 * 48 = 192
Then addition: 192 + 5 = 197
Final answer: 197","math,arithmetic,step-by-step"
arb_1086,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the umbrella from the bedroom and deliver it to the kitchen. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the umbrella (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the kitchen.
6. Deliver / place the umbrella in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1087,commonsense_reasoning,everyday,easy,A purple wolf is observed in the server room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1088,tool_use,selection,easy,"Available tools: database_query, calendar, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1089,commonsense_reasoning,everyday,easy,A purple owl is observed in the server room. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1090,commonsense_reasoning,everyday,easy,A green rabbit is observed in the office. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a office (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1091,commonsense_reasoning,everyday,easy,A green bird is observed in the garage. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1092,commonsense_reasoning,everyday,easy,A purple dog is observed in the lobby. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1093,logical_reasoning,deduction,medium,All Rileys own a notebook. Quinn is a Riley. Does Quinn own a notebook? Explain step by step.,"Yes. From the universal statement 'All Rileys own a notebook' and the particular 'Quinn is a Riley', it follows by universal instantiation that Quinn owns a notebook.","deduction,syllogism,logic"
arb_1094,commonsense_reasoning,everyday,easy,A purple fox is observed in the meeting room. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1095,logical_reasoning,deduction,easy,"Either Taylor or Jordan took the book, but not both. Taylor did not take the book. Who took it?",Jordan took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1096,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1097,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the book from the kitchen and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the book (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the library.
6. Deliver / place the book in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1098,commonsense_reasoning,everyday,easy,A blue wolf is observed in the office. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1099,logical_reasoning,deduction,easy,No Caseys are in the office. Cameron is in the office. Is Cameron a Casey? Explain.,"No. The premise states that no Caseys are in the office. Since Cameron is in the office, Cameron cannot be a Casey.","deduction,syllogism,logic"
arb_1100,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1101,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1102,mathematical_reasoning,arithmetic,easy,Compute (17 * 38) + 32. Explain the order of operations.,"Multiplication first: 17 * 38 = 646
Then addition: 646 + 32 = 678
Final answer: 678","math,arithmetic,step-by-step"
arb_1103,mathematical_reasoning,arithmetic,easy,Start with 14. Subtract 38. Then add 33. What is the final value? Show all steps.,"Step 1: 14 - 38 = -24
Step 2: -24 + 33 = 9
Final answer: 9","math,arithmetic,step-by-step"
arb_1104,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the notebook from the lobby and deliver it to the warehouse. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the notebook (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the warehouse.
6. Deliver / place the notebook in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1105,logical_reasoning,deduction,medium,"If someone owns a umbrella, then they have access to the kitchen. Riley owns a umbrella. What can we conclude about Riley?","We can conclude that Riley has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1106,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1107,mathematical_reasoning,arithmetic,easy,Start with 47. Add 27. Then multiply the result by 11. What is the final value? Show all steps.,"Step 1: 47 + 27 = 74
Step 2: 74 * 11 = 814
Final answer: 814","math,arithmetic,step-by-step"
arb_1108,tool_use,selection,medium,"Available tools: code_interpreter, email_client, file_system, browser.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1109,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the laptop from the library and deliver it to the warehouse. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the laptop (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1110,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1111,logical_reasoning,deduction,easy,No Camerons are in the office. Avery is in the office. Is Avery a Cameron? Explain.,"No. The premise states that no Camerons are in the office. Since Avery is in the office, Avery cannot be a Cameron.","deduction,syllogism,logic"
arb_1112,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1113,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the laptop from the lab and deliver it to the warehouse. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1114,tool_use,selection,medium,"Available tools: file_system, web_search, calendar, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1115,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1116,mathematical_reasoning,arithmetic,easy,Start with 44. Subtract 22. Then add 41. What is the final value? Show all steps.,"Step 1: 44 - 22 = 22
Step 2: 22 + 41 = 63
Final answer: 63","math,arithmetic,step-by-step"
arb_1117,commonsense_reasoning,everyday,easy,A red rabbit is observed in the lab. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1118,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1119,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1120,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_1121,mathematical_reasoning,arithmetic,easy,Compute (13 * 16) + 49. Explain the order of operations.,"Multiplication first: 13 * 16 = 208
Then addition: 208 + 49 = 257
Final answer: 257","math,arithmetic,step-by-step"
arb_1122,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1123,commonsense_reasoning,everyday,easy,A purple cat is observed in the kitchen. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1124,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the laptop from the office and deliver it to the bedroom. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the bedroom.
6. Deliver / place the laptop in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1125,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1126,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the charger from the server room and deliver it to the kitchen. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the charger (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the kitchen.
6. Deliver / place the charger in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1127,commonsense_reasoning,everyday,easy,A purple fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1128,mathematical_reasoning,arithmetic,easy,Start with 29. Subtract 10. Then add 47. What is the final value? Show all steps.,"Step 1: 29 - 10 = 19
Step 2: 19 + 47 = 66
Final answer: 66","math,arithmetic,step-by-step"
arb_1129,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1130,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1131,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1132,logical_reasoning,deduction,easy,All Caseys own a phone. Morgan is a Casey. Does Morgan own a phone? Explain step by step.,"Yes. From the universal statement 'All Caseys own a phone' and the particular 'Morgan is a Casey', it follows by universal instantiation that Morgan owns a phone.","deduction,syllogism,logic"
arb_1133,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_1134,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1135,commonsense_reasoning,everyday,easy,A red fish is observed in the lab. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1136,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_1137,commonsense_reasoning,everyday,easy,A black fish is observed in the kitchen. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1138,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1139,tool_use,selection,easy,"Available tools: email_client, database_query, code_interpreter, terminal, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1140,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the umbrella from the lobby and deliver it to the garage. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the umbrella (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the garage.
6. Deliver / place the umbrella in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1141,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_1142,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the phone from the garage and deliver it to the meeting room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the phone (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1143,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_1144,tool_use,selection,easy,"Available tools: terminal, file_system, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1145,tool_use,selection,medium,"Available tools: calendar, file_system, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1146,logical_reasoning,deduction,medium,"Either Taylor or Quinn took the notebook, but not both. Taylor did not take the notebook. Who took it?",Quinn took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1147,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the library. Cameron owns a umbrella. What can we conclude about Cameron?","We can conclude that Cameron has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_1148,logical_reasoning,deduction,easy,"Either Avery or Jordan took the laptop, but not both. Avery did not take the laptop. Who took it?",Jordan took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1149,tool_use,selection,medium,"Available tools: terminal, database_query, web_search, file_system, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1150,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1151,logical_reasoning,deduction,easy,"Either Cameron or Avery took the charger, but not both. Cameron did not take the charger. Who took it?",Avery took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1152,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1153,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1154,tool_use,selection,medium,"Available tools: code_interpreter, browser, calendar, file_system.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1155,tool_use,selection,easy,"Available tools: terminal, code_interpreter, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1156,tool_use,selection,easy,"Available tools: calendar, email_client, web_search.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1157,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1158,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1159,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1160,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1161,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1162,mathematical_reasoning,arithmetic,easy,Start with 26. Add 8. Then multiply the result by 1. What is the final value? Show all steps.,"Step 1: 26 + 8 = 34
Step 2: 34 * 1 = 34
Final answer: 34","math,arithmetic,step-by-step"
arb_1163,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the notebook from the warehouse and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the notebook (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the server room.
6. Deliver / place the notebook in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1164,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the lab. Taylor owns a laptop. What can we conclude about Taylor?","We can conclude that Taylor has access to the lab, by modus ponens.","deduction,syllogism,logic"
arb_1165,mathematical_reasoning,arithmetic,easy,Start with 34. Subtract 19. Then add 27. What is the final value? Show all steps.,"Step 1: 34 - 19 = 15
Step 2: 15 + 27 = 42
Final answer: 42","math,arithmetic,step-by-step"
arb_1166,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_1167,mathematical_reasoning,arithmetic,easy,Start with 41. Add 28. Then multiply the result by 6. What is the final value? Show all steps.,"Step 1: 41 + 28 = 69
Step 2: 69 * 6 = 414
Final answer: 414","math,arithmetic,step-by-step"
arb_1168,tool_use,selection,medium,"Available tools: calendar, database_query, file_system.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1169,logical_reasoning,deduction,medium,"Either Taylor or Riley took the phone, but not both. Taylor did not take the phone. Who took it?",Riley took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1170,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1171,tool_use,selection,medium,"Available tools: browser, database_query, email_client, code_interpreter.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1172,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1173,mathematical_reasoning,arithmetic,easy,Compute (31 * 9) + 21. Explain the order of operations.,"Multiplication first: 31 * 9 = 279
Then addition: 279 + 21 = 300
Final answer: 300","math,arithmetic,step-by-step"
arb_1174,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1175,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1176,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_1177,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1178,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1179,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_1180,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1181,logical_reasoning,deduction,easy,"If someone owns a notebook, then they have access to the library. Cameron owns a notebook. What can we conclude about Cameron?","We can conclude that Cameron has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_1182,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1183,tool_use,selection,easy,"Available tools: database_query, terminal, email_client, browser, file_system.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1184,mathematical_reasoning,arithmetic,easy,Start with 23. Subtract 43. Then add 7. What is the final value? Show all steps.,"Step 1: 23 - 43 = -20
Step 2: -20 + 7 = -13
Final answer: -13","math,arithmetic,step-by-step"
arb_1185,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the book from the garage and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the book (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the kitchen.
6. Deliver / place the book in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1186,mathematical_reasoning,arithmetic,easy,Compute (49 * 8) + 7. Explain the order of operations.,"Multiplication first: 49 * 8 = 392
Then addition: 392 + 7 = 399
Final answer: 399","math,arithmetic,step-by-step"
arb_1187,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1188,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1189,mathematical_reasoning,arithmetic,easy,Start with 4. Add 22. Then multiply the result by 34. What is the final value? Show all steps.,"Step 1: 4 + 22 = 26
Step 2: 26 * 34 = 884
Final answer: 884","math,arithmetic,step-by-step"
arb_1190,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1191,mathematical_reasoning,arithmetic,easy,Start with 26. Add 45. Then multiply the result by 8. What is the final value? Show all steps.,"Step 1: 26 + 45 = 71
Step 2: 71 * 8 = 568
Final answer: 568","math,arithmetic,step-by-step"
arb_1192,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the pen from the kitchen and deliver it to the garage. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the pen (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the garage.
6. Deliver / place the pen in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1193,tool_use,selection,medium,"Available tools: database_query, terminal, calendar, browser, email_client.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1194,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1195,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1196,tool_use,selection,medium,"Available tools: database_query, email_client, calendar, browser.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1197,commonsense_reasoning,everyday,easy,A blue fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1198,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1199,logical_reasoning,deduction,easy,All Jordans own a badge. Cameron is a Jordan. Does Cameron own a badge? Explain step by step.,"Yes. From the universal statement 'All Jordans own a badge' and the particular 'Cameron is a Jordan', it follows by universal instantiation that Cameron owns a badge.","deduction,syllogism,logic"
arb_1200,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_1201,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the badge from the server room and deliver it to the bedroom. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the bedroom.
6. Deliver / place the badge in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1202,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1203,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the notebook from the garage and deliver it to the server room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the notebook (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the server room.
6. Deliver / place the notebook in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1204,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1205,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_1206,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1207,tool_use,selection,medium,"Available tools: code_interpreter, file_system, database_query, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1208,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1209,mathematical_reasoning,arithmetic,easy,Start with 49. Subtract 17. Then add 34. What is the final value? Show all steps.,"Step 1: 49 - 17 = 32
Step 2: 32 + 34 = 66
Final answer: 66","math,arithmetic,step-by-step"
arb_1210,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1211,mathematical_reasoning,arithmetic,easy,Compute (40 * 24) + 45. Explain the order of operations.,"Multiplication first: 40 * 24 = 960
Then addition: 960 + 45 = 1005
Final answer: 1005","math,arithmetic,step-by-step"
arb_1212,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1213,tool_use,selection,medium,"Available tools: calendar, file_system, code_interpreter, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1214,tool_use,selection,easy,"Available tools: terminal, code_interpreter, web_search.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1215,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1216,tool_use,selection,easy,"Available tools: web_search, calendar, code_interpreter, terminal, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1217,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1218,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1219,commonsense_reasoning,everyday,easy,A yellow rabbit is observed in the lobby. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1220,mathematical_reasoning,arithmetic,easy,Start with 20. Subtract 16. Then add 34. What is the final value? Show all steps.,"Step 1: 20 - 16 = 4
Step 2: 4 + 34 = 38
Final answer: 38","math,arithmetic,step-by-step"
arb_1221,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1222,logical_reasoning,deduction,medium,All Jordans own a umbrella. Avery is a Jordan. Does Avery own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Jordans own a umbrella' and the particular 'Avery is a Jordan', it follows by universal instantiation that Avery owns a umbrella.","deduction,syllogism,logic"
arb_1223,mathematical_reasoning,arithmetic,easy,Start with 22. Add 21. Then multiply the result by 26. What is the final value? Show all steps.,"Step 1: 22 + 21 = 43
Step 2: 43 * 26 = 1118
Final answer: 1118","math,arithmetic,step-by-step"
arb_1224,tool_use,selection,easy,"Available tools: web_search, browser, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1225,tool_use,selection,medium,"Available tools: email_client, database_query, file_system, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1226,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the notebook from the meeting room and deliver it to the server room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the notebook (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the server room.
6. Deliver / place the notebook in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1227,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1228,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the phone from the server room and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1229,tool_use,selection,easy,"Available tools: web_search, calendar, terminal, file_system, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1230,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_1231,tool_use,selection,easy,"Available tools: code_interpreter, email_client, calendar, browser.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1232,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1233,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the charger from the lab and deliver it to the meeting room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the charger (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the meeting room.
6. Deliver / place the charger in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1234,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the umbrella from the bedroom and deliver it to the lobby. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the umbrella (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the lobby.
6. Deliver / place the umbrella in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1235,commonsense_reasoning,everyday,easy,A white fish is observed in the bedroom. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1236,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_1237,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the notebook from the warehouse and deliver it to the garage. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the notebook (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the garage.
6. Deliver / place the notebook in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1238,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the pen from the library and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the pen (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1239,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1240,tool_use,selection,easy,"Available tools: browser, database_query, code_interpreter.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1241,mathematical_reasoning,arithmetic,easy,Start with 13. Subtract 6. Then add 27. What is the final value? Show all steps.,"Step 1: 13 - 6 = 7
Step 2: 7 + 27 = 34
Final answer: 34","math,arithmetic,step-by-step"
arb_1242,mathematical_reasoning,arithmetic,easy,Start with 28. Subtract 30. Then add 44. What is the final value? Show all steps.,"Step 1: 28 - 30 = -2
Step 2: -2 + 44 = 42
Final answer: 42","math,arithmetic,step-by-step"
arb_1243,tool_use,selection,medium,"Available tools: browser, file_system, calendar.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1244,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1245,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1246,commonsense_reasoning,everyday,easy,A blue owl is observed in the office. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a office (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1247,mathematical_reasoning,arithmetic,easy,Compute (32 * 49) + 10. Explain the order of operations.,"Multiplication first: 32 * 49 = 1568
Then addition: 1568 + 10 = 1578
Final answer: 1578","math,arithmetic,step-by-step"
arb_1248,tool_use,selection,medium,"Available tools: web_search, email_client, browser, calendar, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1249,mathematical_reasoning,arithmetic,easy,Compute (41 * 18) + 39. Explain the order of operations.,"Multiplication first: 41 * 18 = 738
Then addition: 738 + 39 = 777
Final answer: 777","math,arithmetic,step-by-step"
arb_1250,mathematical_reasoning,arithmetic,easy,Start with 32. Subtract 31. Then add 2. What is the final value? Show all steps.,"Step 1: 32 - 31 = 1
Step 2: 1 + 2 = 3
Final answer: 3","math,arithmetic,step-by-step"
arb_1251,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1252,mathematical_reasoning,arithmetic,easy,Compute (47 * 30) + 49. Explain the order of operations.,"Multiplication first: 47 * 30 = 1410
Then addition: 1410 + 49 = 1459
Final answer: 1459","math,arithmetic,step-by-step"
arb_1253,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1254,mathematical_reasoning,arithmetic,easy,Compute (36 * 6) + 13. Explain the order of operations.,"Multiplication first: 36 * 6 = 216
Then addition: 216 + 13 = 229
Final answer: 229","math,arithmetic,step-by-step"
arb_1255,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1256,mathematical_reasoning,arithmetic,easy,Compute (3 * 25) + 21. Explain the order of operations.,"Multiplication first: 3 * 25 = 75
Then addition: 75 + 21 = 96
Final answer: 96","math,arithmetic,step-by-step"
arb_1257,commonsense_reasoning,everyday,easy,A green wolf is observed in the library. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1258,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_1259,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_1260,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1261,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1262,tool_use,selection,medium,"Available tools: code_interpreter, email_client, database_query, web_search, calendar.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1263,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1264,tool_use,selection,medium,"Available tools: code_interpreter, browser, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1265,mathematical_reasoning,arithmetic,easy,Start with 35. Add 37. Then multiply the result by 49. What is the final value? Show all steps.,"Step 1: 35 + 37 = 72
Step 2: 72 * 49 = 3528
Final answer: 3528","math,arithmetic,step-by-step"
arb_1266,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1267,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the pen from the office and deliver it to the server room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1268,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1269,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1270,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1271,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1272,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1273,commonsense_reasoning,everyday,easy,A white rabbit is observed in the lab. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1274,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1275,tool_use,selection,medium,"Available tools: calendar, browser, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1276,logical_reasoning,deduction,medium,"Either Cameron or Morgan took the book, but not both. Cameron did not take the book. Who took it?",Morgan took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1277,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1278,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_1279,commonsense_reasoning,everyday,easy,A white rabbit is observed in the lab. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1280,tool_use,selection,easy,"Available tools: browser, code_interpreter, calendar, terminal, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1281,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1282,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1283,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the wallet from the bedroom and deliver it to the library. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the wallet (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the library.
6. Deliver / place the wallet in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1284,tool_use,selection,easy,"Available tools: web_search, file_system, calendar, terminal.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1285,commonsense_reasoning,everyday,easy,A blue wolf is observed in the kitchen. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1286,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the key from the lobby and deliver it to the kitchen. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the kitchen.
6. Deliver / place the key in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1287,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1288,logical_reasoning,deduction,medium,No Alexs are in the garage. Casey is in the garage. Is Casey a Alex? Explain.,"No. The premise states that no Alexs are in the garage. Since Casey is in the garage, Casey cannot be a Alex.","deduction,syllogism,logic"
arb_1289,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the phone from the lab and deliver it to the server room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the server room.
6. Deliver / place the phone in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1290,tool_use,selection,medium,"Available tools: code_interpreter, calendar, web_search, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1291,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1292,tool_use,selection,easy,"Available tools: file_system, email_client, calendar.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1293,tool_use,selection,medium,"Available tools: calendar, code_interpreter, browser.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1294,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1295,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the notebook from the garage and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the notebook (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the server room.
6. Deliver / place the notebook in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1296,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the book from the meeting room and deliver it to the office. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the book (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the office.
6. Deliver / place the book in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1297,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the pen from the kitchen and deliver it to the bedroom. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the pen (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the bedroom.
6. Deliver / place the pen in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1298,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1299,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1300,logical_reasoning,deduction,medium,"Either Casey or Jordan took the notebook, but not both. Casey did not take the notebook. Who took it?",Jordan took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1301,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1302,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1303,logical_reasoning,deduction,easy,No Morgans are in the garage. Jamie is in the garage. Is Jamie a Morgan? Explain.,"No. The premise states that no Morgans are in the garage. Since Jamie is in the garage, Jamie cannot be a Morgan.","deduction,syllogism,logic"
arb_1304,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_1305,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_1306,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1307,mathematical_reasoning,arithmetic,easy,Start with 46. Add 34. Then multiply the result by 5. What is the final value? Show all steps.,"Step 1: 46 + 34 = 80
Step 2: 80 * 5 = 400
Final answer: 400","math,arithmetic,step-by-step"
arb_1308,logical_reasoning,deduction,medium,"Either Morgan or Jordan took the pen, but not both. Morgan did not take the pen. Who took it?",Jordan took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1309,mathematical_reasoning,arithmetic,easy,Start with 12. Subtract 3. Then add 31. What is the final value? Show all steps.,"Step 1: 12 - 3 = 9
Step 2: 9 + 31 = 40
Final answer: 40","math,arithmetic,step-by-step"
arb_1310,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1311,logical_reasoning,deduction,easy,"Either Taylor or Quinn took the pen, but not both. Taylor did not take the pen. Who took it?",Quinn took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1312,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the book from the server room and deliver it to the garage. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the book (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the garage.
6. Deliver / place the book in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1313,commonsense_reasoning,everyday,easy,A white fox is observed in the lab. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1314,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1315,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1316,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_1317,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_1318,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the laptop from the warehouse and deliver it to the server room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the laptop (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the server room.
6. Deliver / place the laptop in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1319,mathematical_reasoning,arithmetic,easy,Start with 5. Subtract 18. Then add 42. What is the final value? Show all steps.,"Step 1: 5 - 18 = -13
Step 2: -13 + 42 = 29
Final answer: 29","math,arithmetic,step-by-step"
arb_1320,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_1321,tool_use,selection,easy,"Available tools: terminal, calendar, file_system.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1322,commonsense_reasoning,everyday,easy,A blue cat is observed in the warehouse. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1323,logical_reasoning,deduction,medium,All Rileys own a pen. Quinn is a Riley. Does Quinn own a pen? Explain step by step.,"Yes. From the universal statement 'All Rileys own a pen' and the particular 'Quinn is a Riley', it follows by universal instantiation that Quinn owns a pen.","deduction,syllogism,logic"
arb_1324,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1325,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the key from the office and deliver it to the lab. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the key (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lab.
6. Deliver / place the key in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1326,logical_reasoning,deduction,easy,"If someone owns a book, then they have access to the kitchen. Taylor owns a book. What can we conclude about Taylor?","We can conclude that Taylor has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1327,mathematical_reasoning,arithmetic,easy,Start with 31. Add 14. Then multiply the result by 29. What is the final value? Show all steps.,"Step 1: 31 + 14 = 45
Step 2: 45 * 29 = 1305
Final answer: 1305","math,arithmetic,step-by-step"
arb_1328,logical_reasoning,deduction,medium,All Jamies own a phone. Avery is a Jamie. Does Avery own a phone? Explain step by step.,"Yes. From the universal statement 'All Jamies own a phone' and the particular 'Avery is a Jamie', it follows by universal instantiation that Avery owns a phone.","deduction,syllogism,logic"
arb_1329,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1330,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1331,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1332,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1333,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1334,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the key from the garage and deliver it to the meeting room. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the key (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1335,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the wallet from the meeting room and deliver it to the garage. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1336,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the laptop from the bedroom and deliver it to the library. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the library.
6. Deliver / place the laptop in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1337,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the charger from the bedroom and deliver it to the garage. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1338,commonsense_reasoning,everyday,easy,A blue cat is observed in the warehouse. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1339,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1340,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_1341,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_1342,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1343,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the notebook from the library and deliver it to the meeting room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the notebook (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1344,logical_reasoning,deduction,easy,"Either Alex or Taylor took the book, but not both. Alex did not take the book. Who took it?",Taylor took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1345,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1346,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1347,tool_use,selection,easy,"Available tools: database_query, file_system, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1348,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the charger from the lab and deliver it to the warehouse. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the warehouse.
6. Deliver / place the charger in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1349,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1350,mathematical_reasoning,arithmetic,easy,Compute (12 * 27) + 2. Explain the order of operations.,"Multiplication first: 12 * 27 = 324
Then addition: 324 + 2 = 326
Final answer: 326","math,arithmetic,step-by-step"
arb_1351,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1352,tool_use,selection,medium,"Available tools: code_interpreter, calendar, terminal, file_system, web_search.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1353,logical_reasoning,deduction,medium,All Quinns own a notebook. Alex is a Quinn. Does Alex own a notebook? Explain step by step.,"Yes. From the universal statement 'All Quinns own a notebook' and the particular 'Alex is a Quinn', it follows by universal instantiation that Alex owns a notebook.","deduction,syllogism,logic"
arb_1354,commonsense_reasoning,everyday,easy,A black fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1355,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1356,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1357,logical_reasoning,deduction,easy,"If someone owns a key, then they have access to the garage. Jamie owns a key. What can we conclude about Jamie?","We can conclude that Jamie has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_1358,commonsense_reasoning,everyday,easy,A yellow fish is observed in the server room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1359,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1360,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the laptop from the server room and deliver it to the office. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the laptop (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the office.
6. Deliver / place the laptop in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1361,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the pen from the bedroom and deliver it to the meeting room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the pen (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the meeting room.
6. Deliver / place the pen in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1362,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1363,mathematical_reasoning,arithmetic,easy,Compute (29 * 1) + 15. Explain the order of operations.,"Multiplication first: 29 * 1 = 29
Then addition: 29 + 15 = 44
Final answer: 44","math,arithmetic,step-by-step"
arb_1364,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1365,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the charger from the lobby and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the charger (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the kitchen.
6. Deliver / place the charger in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1366,logical_reasoning,deduction,medium,No Alexs are in the bedroom. Cameron is in the bedroom. Is Cameron a Alex? Explain.,"No. The premise states that no Alexs are in the bedroom. Since Cameron is in the bedroom, Cameron cannot be a Alex.","deduction,syllogism,logic"
arb_1367,logical_reasoning,deduction,easy,No Camerons are in the kitchen. Alex is in the kitchen. Is Alex a Cameron? Explain.,"No. The premise states that no Camerons are in the kitchen. Since Alex is in the kitchen, Alex cannot be a Cameron.","deduction,syllogism,logic"
arb_1368,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1369,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_1370,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the charger from the garage and deliver it to the library. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the charger (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the library.
6. Deliver / place the charger in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1371,mathematical_reasoning,arithmetic,easy,Start with 23. Subtract 25. Then add 39. What is the final value? Show all steps.,"Step 1: 23 - 25 = -2
Step 2: -2 + 39 = 37
Final answer: 37","math,arithmetic,step-by-step"
arb_1372,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the umbrella from the garage and deliver it to the lobby. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the umbrella (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the lobby.
6. Deliver / place the umbrella in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1373,tool_use,selection,easy,"Available tools: database_query, calendar, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1374,commonsense_reasoning,everyday,easy,A green owl is observed in the bedroom. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1375,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_1376,logical_reasoning,deduction,easy,"Either Cameron or Alex took the notebook, but not both. Cameron did not take the notebook. Who took it?",Alex took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1377,tool_use,selection,medium,"Available tools: terminal, email_client, browser, web_search, file_system.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1378,mathematical_reasoning,arithmetic,easy,Start with 8. Add 2. Then multiply the result by 45. What is the final value? Show all steps.,"Step 1: 8 + 2 = 10
Step 2: 10 * 45 = 450
Final answer: 450","math,arithmetic,step-by-step"
arb_1379,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the wallet from the meeting room and deliver it to the bedroom. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the wallet (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the bedroom.
6. Deliver / place the wallet in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1380,commonsense_reasoning,everyday,easy,A white fish is observed in the warehouse. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1381,mathematical_reasoning,arithmetic,easy,Start with 10. Add 8. Then multiply the result by 23. What is the final value? Show all steps.,"Step 1: 10 + 8 = 18
Step 2: 18 * 23 = 414
Final answer: 414","math,arithmetic,step-by-step"
arb_1382,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_1383,commonsense_reasoning,everyday,easy,A green wolf is observed in the office. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1384,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the pen from the warehouse and deliver it to the garage. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the pen (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the garage.
6. Deliver / place the pen in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1385,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the phone from the garage and deliver it to the lobby. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the phone (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the lobby.
6. Deliver / place the phone in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1386,logical_reasoning,deduction,medium,All Quinns own a pen. Taylor is a Quinn. Does Taylor own a pen? Explain step by step.,"Yes. From the universal statement 'All Quinns own a pen' and the particular 'Taylor is a Quinn', it follows by universal instantiation that Taylor owns a pen.","deduction,syllogism,logic"
arb_1387,logical_reasoning,deduction,easy,No Taylors are in the lobby. Avery is in the lobby. Is Avery a Taylor? Explain.,"No. The premise states that no Taylors are in the lobby. Since Avery is in the lobby, Avery cannot be a Taylor.","deduction,syllogism,logic"
arb_1388,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_1389,tool_use,selection,easy,"Available tools: terminal, web_search, calendar.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1390,commonsense_reasoning,everyday,easy,A green cat is observed in the office. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a office (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1391,logical_reasoning,deduction,easy,"Either Jamie or Alex took the book, but not both. Jamie did not take the book. Who took it?",Alex took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1392,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1393,mathematical_reasoning,arithmetic,easy,Start with 27. Subtract 21. Then add 29. What is the final value? Show all steps.,"Step 1: 27 - 21 = 6
Step 2: 6 + 29 = 35
Final answer: 35","math,arithmetic,step-by-step"
arb_1394,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1395,mathematical_reasoning,arithmetic,easy,Start with 41. Add 35. Then multiply the result by 13. What is the final value? Show all steps.,"Step 1: 41 + 35 = 76
Step 2: 76 * 13 = 988
Final answer: 988","math,arithmetic,step-by-step"
arb_1396,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1397,tool_use,selection,easy,"Available tools: file_system, code_interpreter, browser.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1398,mathematical_reasoning,arithmetic,easy,Start with 33. Subtract 32. Then add 22. What is the final value? Show all steps.,"Step 1: 33 - 32 = 1
Step 2: 1 + 22 = 23
Final answer: 23","math,arithmetic,step-by-step"
arb_1399,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the laptop from the lab and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the laptop (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the kitchen.
6. Deliver / place the laptop in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1400,mathematical_reasoning,arithmetic,easy,Start with 38. Add 8. Then multiply the result by 6. What is the final value? Show all steps.,"Step 1: 38 + 8 = 46
Step 2: 46 * 6 = 276
Final answer: 276","math,arithmetic,step-by-step"
arb_1401,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1402,mathematical_reasoning,arithmetic,easy,Compute (20 * 32) + 28. Explain the order of operations.,"Multiplication first: 20 * 32 = 640
Then addition: 640 + 28 = 668
Final answer: 668","math,arithmetic,step-by-step"
arb_1403,logical_reasoning,deduction,medium,No Caseys are in the lab. Alex is in the lab. Is Alex a Casey? Explain.,"No. The premise states that no Caseys are in the lab. Since Alex is in the lab, Alex cannot be a Casey.","deduction,syllogism,logic"
arb_1404,commonsense_reasoning,everyday,easy,A green bird is observed in the kitchen. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1405,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1406,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1407,commonsense_reasoning,everyday,easy,A green owl is observed in the library. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a library (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1408,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the pen from the office and deliver it to the server room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1409,commonsense_reasoning,everyday,easy,A blue cat is observed in the server room. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1410,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the umbrella from the lobby and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the umbrella (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the meeting room.
6. Deliver / place the umbrella in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1411,mathematical_reasoning,arithmetic,easy,Start with 33. Add 39. Then multiply the result by 43. What is the final value? Show all steps.,"Step 1: 33 + 39 = 72
Step 2: 72 * 43 = 3096
Final answer: 3096","math,arithmetic,step-by-step"
arb_1412,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the wallet from the meeting room and deliver it to the garage. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the wallet (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1413,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the wallet from the kitchen and deliver it to the warehouse. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the wallet (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the warehouse.
6. Deliver / place the wallet in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1414,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1415,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1416,commonsense_reasoning,everyday,easy,A white wolf is observed in the warehouse. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1417,mathematical_reasoning,arithmetic,easy,Start with 23. Subtract 46. Then add 36. What is the final value? Show all steps.,"Step 1: 23 - 46 = -23
Step 2: -23 + 36 = 13
Final answer: 13","math,arithmetic,step-by-step"
arb_1418,logical_reasoning,deduction,easy,"Either Avery or Cameron took the phone, but not both. Avery did not take the phone. Who took it?",Cameron took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1419,commonsense_reasoning,everyday,easy,A blue fox is observed in the kitchen. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1420,mathematical_reasoning,arithmetic,easy,Start with 8. Add 30. Then multiply the result by 24. What is the final value? Show all steps.,"Step 1: 8 + 30 = 38
Step 2: 38 * 24 = 912
Final answer: 912","math,arithmetic,step-by-step"
arb_1421,logical_reasoning,deduction,easy,"If someone owns a phone, then they have access to the garage. Alex owns a phone. What can we conclude about Alex?","We can conclude that Alex has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_1422,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1423,tool_use,selection,easy,"Available tools: calendar, web_search, browser, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1424,mathematical_reasoning,arithmetic,easy,Start with 25. Add 12. Then multiply the result by 45. What is the final value? Show all steps.,"Step 1: 25 + 12 = 37
Step 2: 37 * 45 = 1665
Final answer: 1665","math,arithmetic,step-by-step"
arb_1425,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the charger from the office and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the charger (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the meeting room.
6. Deliver / place the charger in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1426,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1427,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1428,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the charger from the garage and deliver it to the bedroom. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the bedroom.
6. Deliver / place the charger in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1429,logical_reasoning,deduction,medium,"If someone owns a pen, then they have access to the warehouse. Jamie owns a pen. What can we conclude about Jamie?","We can conclude that Jamie has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_1430,tool_use,selection,easy,"Available tools: email_client, file_system, web_search, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1431,tool_use,selection,medium,"Available tools: email_client, web_search, calendar.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1432,tool_use,selection,medium,"Available tools: browser, terminal, calendar, database_query, code_interpreter.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1433,logical_reasoning,deduction,easy,No Jordans are in the office. Morgan is in the office. Is Morgan a Jordan? Explain.,"No. The premise states that no Jordans are in the office. Since Morgan is in the office, Morgan cannot be a Jordan.","deduction,syllogism,logic"
arb_1434,commonsense_reasoning,everyday,easy,A purple cat is observed in the kitchen. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1435,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the library. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a library (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1436,mathematical_reasoning,arithmetic,easy,Start with 2. Subtract 35. Then add 38. What is the final value? Show all steps.,"Step 1: 2 - 35 = -33
Step 2: -33 + 38 = 5
Final answer: 5","math,arithmetic,step-by-step"
arb_1437,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the charger from the bedroom and deliver it to the garage. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1438,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1439,commonsense_reasoning,everyday,easy,A green cat is observed in the meeting room. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1440,commonsense_reasoning,everyday,easy,A white wolf is observed in the library. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1441,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1442,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1443,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the book from the server room and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the book (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the office.
6. Deliver / place the book in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1444,commonsense_reasoning,everyday,easy,A purple owl is observed in the warehouse. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1445,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1446,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_1447,commonsense_reasoning,everyday,easy,A green cat is observed in the bedroom. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1448,logical_reasoning,deduction,medium,"If someone owns a badge, then they have access to the library. Jordan owns a badge. What can we conclude about Jordan?","We can conclude that Jordan has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_1449,tool_use,selection,easy,"Available tools: calendar, code_interpreter, terminal, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1450,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1451,tool_use,selection,medium,"Available tools: code_interpreter, web_search, file_system, database_query.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1452,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1453,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1454,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1455,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1456,tool_use,selection,medium,"Available tools: file_system, email_client, database_query, calendar, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1457,logical_reasoning,deduction,easy,"If someone owns a phone, then they have access to the lab. Cameron owns a phone. What can we conclude about Cameron?","We can conclude that Cameron has access to the lab, by modus ponens.","deduction,syllogism,logic"
arb_1458,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1459,mathematical_reasoning,arithmetic,easy,Start with 49. Add 42. Then multiply the result by 10. What is the final value? Show all steps.,"Step 1: 49 + 42 = 91
Step 2: 91 * 10 = 910
Final answer: 910","math,arithmetic,step-by-step"
arb_1460,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1461,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_1462,mathematical_reasoning,arithmetic,easy,Compute (12 * 26) + 25. Explain the order of operations.,"Multiplication first: 12 * 26 = 312
Then addition: 312 + 25 = 337
Final answer: 337","math,arithmetic,step-by-step"
arb_1463,logical_reasoning,deduction,easy,"If someone owns a phone, then they have access to the bedroom. Avery owns a phone. What can we conclude about Avery?","We can conclude that Avery has access to the bedroom, by modus ponens.","deduction,syllogism,logic"
arb_1464,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1465,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1466,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the phone from the meeting room and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the phone (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the server room.
6. Deliver / place the phone in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1467,tool_use,selection,easy,"Available tools: file_system, email_client, database_query, terminal, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1468,mathematical_reasoning,arithmetic,easy,Start with 34. Add 8. Then multiply the result by 45. What is the final value? Show all steps.,"Step 1: 34 + 8 = 42
Step 2: 42 * 45 = 1890
Final answer: 1890","math,arithmetic,step-by-step"
arb_1469,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1470,tool_use,selection,medium,"Available tools: terminal, file_system, calendar, browser, email_client.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1471,commonsense_reasoning,everyday,easy,A blue fish is observed in the kitchen. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1472,logical_reasoning,deduction,medium,No Caseys are in the bedroom. Jordan is in the bedroom. Is Jordan a Casey? Explain.,"No. The premise states that no Caseys are in the bedroom. Since Jordan is in the bedroom, Jordan cannot be a Casey.","deduction,syllogism,logic"
arb_1473,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_1474,commonsense_reasoning,everyday,easy,A black bird is observed in the meeting room. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1475,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the office. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a office (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1476,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1477,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the charger from the warehouse and deliver it to the lab. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1478,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the key from the lobby and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1479,commonsense_reasoning,everyday,easy,A green bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1480,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_1481,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1482,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the laptop from the lobby and deliver it to the server room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the laptop (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the server room.
6. Deliver / place the laptop in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1483,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1484,tool_use,selection,medium,"Available tools: web_search, email_client, calendar, terminal, database_query.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1485,commonsense_reasoning,everyday,easy,A red bird is observed in the meeting room. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1486,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the data loss.","causal,diagnosis,agentic"
arb_1487,commonsense_reasoning,everyday,easy,A red wolf is observed in the library. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1488,commonsense_reasoning,everyday,easy,A orange rabbit is observed in the library. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a library (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1489,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1490,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1491,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1492,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1493,mathematical_reasoning,arithmetic,easy,Compute (48 * 33) + 18. Explain the order of operations.,"Multiplication first: 48 * 33 = 1584
Then addition: 1584 + 18 = 1602
Final answer: 1602","math,arithmetic,step-by-step"
arb_1494,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1495,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the error logs.","causal,diagnosis,agentic"
arb_1496,logical_reasoning,deduction,easy,"Either Alex or Quinn took the phone, but not both. Alex did not take the phone. Who took it?",Quinn took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1497,mathematical_reasoning,arithmetic,easy,Start with 1. Add 39. Then multiply the result by 31. What is the final value? Show all steps.,"Step 1: 1 + 39 = 40
Step 2: 40 * 31 = 1240
Final answer: 1240","math,arithmetic,step-by-step"
arb_1498,logical_reasoning,deduction,medium,No Quinns are in the server room. Alex is in the server room. Is Alex a Quinn? Explain.,"No. The premise states that no Quinns are in the server room. Since Alex is in the server room, Alex cannot be a Quinn.","deduction,syllogism,logic"
arb_1499,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1500,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1501,tool_use,selection,easy,"Available tools: email_client, web_search, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1502,mathematical_reasoning,arithmetic,easy,Start with 23. Add 42. Then multiply the result by 26. What is the final value? Show all steps.,"Step 1: 23 + 42 = 65
Step 2: 65 * 26 = 1690
Final answer: 1690","math,arithmetic,step-by-step"
arb_1503,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1504,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the phone from the meeting room and deliver it to the bedroom. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the phone (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the bedroom.
6. Deliver / place the phone in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1505,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1506,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1507,commonsense_reasoning,everyday,easy,A black bird is observed in the lobby. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1508,commonsense_reasoning,everyday,easy,A blue fox is observed in the lab. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1509,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the charger from the kitchen and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the charger (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1510,logical_reasoning,deduction,medium,No Quinns are in the warehouse. Avery is in the warehouse. Is Avery a Quinn? Explain.,"No. The premise states that no Quinns are in the warehouse. Since Avery is in the warehouse, Avery cannot be a Quinn.","deduction,syllogism,logic"
arb_1511,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the badge from the lab and deliver it to the kitchen. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the badge (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the kitchen.
6. Deliver / place the badge in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1512,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1513,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1514,logical_reasoning,deduction,easy,"Either Riley or Morgan took the laptop, but not both. Riley did not take the laptop. Who took it?",Morgan took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1515,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1516,logical_reasoning,deduction,medium,"If someone owns a umbrella, then they have access to the office. Cameron owns a umbrella. What can we conclude about Cameron?","We can conclude that Cameron has access to the office, by modus ponens.","deduction,syllogism,logic"
arb_1517,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1518,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the badge from the lab and deliver it to the server room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the badge (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the server room.
6. Deliver / place the badge in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1519,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_1520,logical_reasoning,deduction,easy,"Either Riley or Jordan took the phone, but not both. Riley did not take the phone. Who took it?",Jordan took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1521,logical_reasoning,deduction,easy,"If someone owns a wallet, then they have access to the meeting room. Casey owns a wallet. What can we conclude about Casey?","We can conclude that Casey has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_1522,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1523,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1524,commonsense_reasoning,everyday,easy,A purple fox is observed in the lobby. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1525,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the laptop from the server room and deliver it to the lab. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the laptop (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1526,commonsense_reasoning,everyday,easy,A white fish is observed in the warehouse. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1527,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1528,logical_reasoning,deduction,medium,All Camerons own a notebook. Quinn is a Cameron. Does Quinn own a notebook? Explain step by step.,"Yes. From the universal statement 'All Camerons own a notebook' and the particular 'Quinn is a Cameron', it follows by universal instantiation that Quinn owns a notebook.","deduction,syllogism,logic"
arb_1529,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the wallet from the office and deliver it to the kitchen. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the wallet (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the kitchen.
6. Deliver / place the wallet in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1530,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the charger from the warehouse and deliver it to the garage. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1531,logical_reasoning,deduction,easy,"Either Taylor or Riley took the pen, but not both. Taylor did not take the pen. Who took it?",Riley took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1532,commonsense_reasoning,everyday,easy,A black owl is observed in the library. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a library (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1533,tool_use,selection,easy,"Available tools: calendar, code_interpreter, file_system, web_search.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1534,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the badge from the lobby and deliver it to the library. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the badge (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the library.
6. Deliver / place the badge in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1535,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1536,tool_use,selection,medium,"Available tools: web_search, database_query, email_client, file_system.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1537,mathematical_reasoning,arithmetic,easy,Start with 34. Add 48. Then multiply the result by 16. What is the final value? Show all steps.,"Step 1: 34 + 48 = 82
Step 2: 82 * 16 = 1312
Final answer: 1312","math,arithmetic,step-by-step"
arb_1538,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_1539,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1540,logical_reasoning,deduction,easy,No Caseys are in the lab. Jordan is in the lab. Is Jordan a Casey? Explain.,"No. The premise states that no Caseys are in the lab. Since Jordan is in the lab, Jordan cannot be a Casey.","deduction,syllogism,logic"
arb_1541,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1542,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1543,tool_use,selection,easy,"Available tools: calendar, file_system, code_interpreter, web_search, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1544,commonsense_reasoning,everyday,easy,A black fish is observed in the lobby. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1545,tool_use,selection,medium,"Available tools: database_query, code_interpreter, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1546,logical_reasoning,deduction,easy,All Morgans own a pen. Quinn is a Morgan. Does Quinn own a pen? Explain step by step.,"Yes. From the universal statement 'All Morgans own a pen' and the particular 'Quinn is a Morgan', it follows by universal instantiation that Quinn owns a pen.","deduction,syllogism,logic"
arb_1547,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1548,tool_use,selection,easy,"Available tools: email_client, browser, code_interpreter.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1549,tool_use,selection,easy,"Available tools: web_search, terminal, file_system.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1550,logical_reasoning,deduction,medium,"Either Jamie or Quinn took the pen, but not both. Jamie did not take the pen. Who took it?",Quinn took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1551,tool_use,selection,medium,"Available tools: calendar, web_search, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1552,commonsense_reasoning,everyday,easy,A blue owl is observed in the garage. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1553,logical_reasoning,deduction,easy,All Jamies own a book. Cameron is a Jamie. Does Cameron own a book? Explain step by step.,"Yes. From the universal statement 'All Jamies own a book' and the particular 'Cameron is a Jamie', it follows by universal instantiation that Cameron owns a book.","deduction,syllogism,logic"
arb_1554,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1555,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_1556,mathematical_reasoning,arithmetic,easy,Start with 1. Subtract 30. Then add 25. What is the final value? Show all steps.,"Step 1: 1 - 30 = -29
Step 2: -29 + 25 = -4
Final answer: -4","math,arithmetic,step-by-step"
arb_1557,tool_use,selection,easy,"Available tools: browser, terminal, web_search, calendar, email_client.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1558,mathematical_reasoning,arithmetic,easy,Compute (19 * 14) + 6. Explain the order of operations.,"Multiplication first: 19 * 14 = 266
Then addition: 266 + 6 = 272
Final answer: 272","math,arithmetic,step-by-step"
arb_1559,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_1560,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1561,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1562,commonsense_reasoning,everyday,easy,A blue fish is observed in the server room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1563,logical_reasoning,deduction,easy,"If someone owns a pen, then they have access to the server room. Casey owns a pen. What can we conclude about Casey?","We can conclude that Casey has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_1564,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the key from the bedroom and deliver it to the lab. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the key (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lab.
6. Deliver / place the key in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1565,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the badge from the kitchen and deliver it to the library. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the badge (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the library.
6. Deliver / place the badge in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1566,tool_use,selection,medium,"Available tools: file_system, calendar, database_query.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1567,tool_use,selection,easy,"Available tools: web_search, email_client, code_interpreter, database_query.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1568,tool_use,selection,medium,"Available tools: email_client, web_search, browser, calendar.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1569,mathematical_reasoning,arithmetic,easy,Start with 31. Subtract 18. Then add 39. What is the final value? Show all steps.,"Step 1: 31 - 18 = 13
Step 2: 13 + 39 = 52
Final answer: 52","math,arithmetic,step-by-step"
arb_1570,commonsense_reasoning,everyday,easy,A white owl is observed in the bedroom. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1571,tool_use,selection,medium,"Available tools: code_interpreter, calendar, terminal, browser, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1572,tool_use,selection,easy,"Available tools: email_client, terminal, file_system.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1573,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1574,logical_reasoning,deduction,easy,"If someone owns a book, then they have access to the lobby. Taylor owns a book. What can we conclude about Taylor?","We can conclude that Taylor has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_1575,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1576,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_1577,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the pen from the lobby and deliver it to the office. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the pen (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the office.
6. Deliver / place the pen in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1578,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1579,tool_use,selection,easy,"Available tools: code_interpreter, email_client, browser.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1580,mathematical_reasoning,arithmetic,easy,Start with 17. Subtract 12. Then add 16. What is the final value? Show all steps.,"Step 1: 17 - 12 = 5
Step 2: 5 + 16 = 21
Final answer: 21","math,arithmetic,step-by-step"
arb_1581,logical_reasoning,deduction,medium,"Either Alex or Avery took the notebook, but not both. Alex did not take the notebook. Who took it?",Avery took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1582,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1583,commonsense_reasoning,everyday,easy,A white fox is observed in the bedroom. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1584,mathematical_reasoning,arithmetic,easy,Start with 49. Add 34. Then multiply the result by 27. What is the final value? Show all steps.,"Step 1: 49 + 34 = 83
Step 2: 83 * 27 = 2241
Final answer: 2241","math,arithmetic,step-by-step"
arb_1585,tool_use,selection,medium,"Available tools: email_client, terminal, calendar, file_system, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1586,logical_reasoning,deduction,easy,"If someone owns a phone, then they have access to the garage. Morgan owns a phone. What can we conclude about Morgan?","We can conclude that Morgan has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_1587,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1588,logical_reasoning,deduction,medium,No Morgans are in the bedroom. Avery is in the bedroom. Is Avery a Morgan? Explain.,"No. The premise states that no Morgans are in the bedroom. Since Avery is in the bedroom, Avery cannot be a Morgan.","deduction,syllogism,logic"
arb_1589,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_1590,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1591,commonsense_reasoning,everyday,easy,A white cat is observed in the lab. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1592,mathematical_reasoning,arithmetic,easy,Start with 43. Add 30. Then multiply the result by 12. What is the final value? Show all steps.,"Step 1: 43 + 30 = 73
Step 2: 73 * 12 = 876
Final answer: 876","math,arithmetic,step-by-step"
arb_1593,commonsense_reasoning,everyday,easy,A green rabbit is observed in the bedroom. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1594,tool_use,selection,medium,"Available tools: web_search, database_query, code_interpreter, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1595,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the key from the kitchen and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the key (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the garage.
6. Deliver / place the key in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1596,logical_reasoning,deduction,easy,"Either Jamie or Casey took the wallet, but not both. Jamie did not take the wallet. Who took it?",Casey took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1597,mathematical_reasoning,arithmetic,easy,Compute (18 * 32) + 46. Explain the order of operations.,"Multiplication first: 18 * 32 = 576
Then addition: 576 + 46 = 622
Final answer: 622","math,arithmetic,step-by-step"
arb_1598,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_1599,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the laptop from the meeting room and deliver it to the warehouse. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the laptop (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1600,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_1601,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1602,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1603,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1604,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1605,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1606,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1607,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1608,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the pen from the server room and deliver it to the bedroom. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the pen (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the bedroom.
6. Deliver / place the pen in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1609,tool_use,selection,easy,"Available tools: web_search, file_system, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1610,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the wallet from the library and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the meeting room.
6. Deliver / place the wallet in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1611,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1612,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the warehouse. Casey owns a umbrella. What can we conclude about Casey?","We can conclude that Casey has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_1613,mathematical_reasoning,arithmetic,easy,Start with 47. Subtract 36. Then add 18. What is the final value? Show all steps.,"Step 1: 47 - 36 = 11
Step 2: 11 + 18 = 29
Final answer: 29","math,arithmetic,step-by-step"
arb_1614,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the charger from the library and deliver it to the kitchen. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the charger (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the kitchen.
6. Deliver / place the charger in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1615,tool_use,selection,easy,"Available tools: database_query, browser, calendar, file_system.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1616,mathematical_reasoning,arithmetic,easy,Compute (38 * 44) + 22. Explain the order of operations.,"Multiplication first: 38 * 44 = 1672
Then addition: 1672 + 22 = 1694
Final answer: 1694","math,arithmetic,step-by-step"
arb_1617,logical_reasoning,deduction,easy,"Either Alex or Riley took the key, but not both. Alex did not take the key. Who took it?",Riley took the key. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1618,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1619,mathematical_reasoning,arithmetic,easy,Start with 19. Subtract 44. Then add 8. What is the final value? Show all steps.,"Step 1: 19 - 44 = -25
Step 2: -25 + 8 = -17
Final answer: -17","math,arithmetic,step-by-step"
arb_1620,logical_reasoning,deduction,medium,"Either Avery or Jamie took the wallet, but not both. Avery did not take the wallet. Who took it?",Jamie took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1621,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1622,commonsense_reasoning,everyday,easy,A purple bird is observed in the garage. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1623,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the charger from the warehouse and deliver it to the lab. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1624,commonsense_reasoning,everyday,easy,A red owl is observed in the warehouse. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1625,mathematical_reasoning,arithmetic,easy,Compute (24 * 17) + 41. Explain the order of operations.,"Multiplication first: 24 * 17 = 408
Then addition: 408 + 41 = 449
Final answer: 449","math,arithmetic,step-by-step"
arb_1626,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1627,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the laptop from the meeting room and deliver it to the office. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the laptop (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the office.
6. Deliver / place the laptop in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1628,mathematical_reasoning,arithmetic,easy,Start with 37. Subtract 26. Then add 4. What is the final value? Show all steps.,"Step 1: 37 - 26 = 11
Step 2: 11 + 4 = 15
Final answer: 15","math,arithmetic,step-by-step"
arb_1629,mathematical_reasoning,arithmetic,easy,Start with 38. Add 12. Then multiply the result by 17. What is the final value? Show all steps.,"Step 1: 38 + 12 = 50
Step 2: 50 * 17 = 850
Final answer: 850","math,arithmetic,step-by-step"
arb_1630,logical_reasoning,deduction,easy,No Alexs are in the office. Avery is in the office. Is Avery a Alex? Explain.,"No. The premise states that no Alexs are in the office. Since Avery is in the office, Avery cannot be a Alex.","deduction,syllogism,logic"
arb_1631,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the badge from the server room and deliver it to the office. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the office.
6. Deliver / place the badge in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1632,logical_reasoning,deduction,medium,"If someone owns a pen, then they have access to the garage. Morgan owns a pen. What can we conclude about Morgan?","We can conclude that Morgan has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_1633,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1634,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1635,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_1636,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1637,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the pen from the library and deliver it to the kitchen. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the pen (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the kitchen.
6. Deliver / place the pen in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1638,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1639,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the umbrella from the lobby and deliver it to the server room. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the umbrella (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the server room.
6. Deliver / place the umbrella in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1640,logical_reasoning,deduction,medium,All Morgans own a wallet. Casey is a Morgan. Does Casey own a wallet? Explain step by step.,"Yes. From the universal statement 'All Morgans own a wallet' and the particular 'Casey is a Morgan', it follows by universal instantiation that Casey owns a wallet.","deduction,syllogism,logic"
arb_1641,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the pen from the lab and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the pen (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1642,mathematical_reasoning,arithmetic,easy,Start with 11. Subtract 46. Then add 30. What is the final value? Show all steps.,"Step 1: 11 - 46 = -35
Step 2: -35 + 30 = -5
Final answer: -5","math,arithmetic,step-by-step"
arb_1643,mathematical_reasoning,arithmetic,easy,Start with 28. Add 4. Then multiply the result by 26. What is the final value? Show all steps.,"Step 1: 28 + 4 = 32
Step 2: 32 * 26 = 832
Final answer: 832","math,arithmetic,step-by-step"
arb_1644,logical_reasoning,deduction,easy,"Either Taylor or Casey took the phone, but not both. Taylor did not take the phone. Who took it?",Casey took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1645,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the wallet from the office and deliver it to the lab. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lab.
6. Deliver / place the wallet in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1646,tool_use,selection,medium,"Available tools: code_interpreter, browser, email_client, web_search, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1647,logical_reasoning,deduction,easy,"Either Jamie or Quinn took the wallet, but not both. Jamie did not take the wallet. Who took it?",Quinn took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1648,logical_reasoning,deduction,easy,"If someone owns a notebook, then they have access to the server room. Taylor owns a notebook. What can we conclude about Taylor?","We can conclude that Taylor has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_1649,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the phone from the server room and deliver it to the kitchen. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the kitchen.
6. Deliver / place the phone in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1650,mathematical_reasoning,arithmetic,easy,Start with 37. Add 6. Then multiply the result by 8. What is the final value? Show all steps.,"Step 1: 37 + 6 = 43
Step 2: 43 * 8 = 344
Final answer: 344","math,arithmetic,step-by-step"
arb_1651,commonsense_reasoning,everyday,easy,A orange fox is observed in the warehouse. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1652,commonsense_reasoning,everyday,easy,A orange fish is observed in the library. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1653,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1654,mathematical_reasoning,arithmetic,easy,Compute (47 * 32) + 20. Explain the order of operations.,"Multiplication first: 47 * 32 = 1504
Then addition: 1504 + 20 = 1524
Final answer: 1524","math,arithmetic,step-by-step"
arb_1655,logical_reasoning,deduction,easy,"If someone owns a pen, then they have access to the kitchen. Alex owns a pen. What can we conclude about Alex?","We can conclude that Alex has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1656,tool_use,selection,easy,"Available tools: database_query, email_client, code_interpreter, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1657,mathematical_reasoning,arithmetic,easy,Start with 2. Subtract 10. Then add 38. What is the final value? Show all steps.,"Step 1: 2 - 10 = -8
Step 2: -8 + 38 = 30
Final answer: 30","math,arithmetic,step-by-step"
arb_1658,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1659,tool_use,selection,easy,"Available tools: database_query, calendar, browser.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1660,commonsense_reasoning,everyday,easy,A blue dog is observed in the kitchen. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1661,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_1662,logical_reasoning,deduction,medium,"Either Jamie or Casey took the laptop, but not both. Jamie did not take the laptop. Who took it?",Casey took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1663,commonsense_reasoning,everyday,easy,A red fish is observed in the server room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1664,commonsense_reasoning,everyday,easy,A purple rabbit is observed in the library. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a library (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1665,tool_use,selection,easy,"Available tools: calendar, database_query, browser, terminal, code_interpreter.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1666,tool_use,selection,easy,"Available tools: code_interpreter, database_query, calendar, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1667,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1668,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the lobby. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1669,mathematical_reasoning,arithmetic,easy,Start with 6. Add 9. Then multiply the result by 1. What is the final value? Show all steps.,"Step 1: 6 + 9 = 15
Step 2: 15 * 1 = 15
Final answer: 15","math,arithmetic,step-by-step"
arb_1670,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the key from the server room and deliver it to the bedroom. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the bedroom.
6. Deliver / place the key in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1671,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1672,tool_use,selection,easy,"Available tools: terminal, file_system, email_client, browser.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1673,mathematical_reasoning,arithmetic,easy,Start with 17. Subtract 16. Then add 40. What is the final value? Show all steps.,"Step 1: 17 - 16 = 1
Step 2: 1 + 40 = 41
Final answer: 41","math,arithmetic,step-by-step"
arb_1674,mathematical_reasoning,arithmetic,easy,Start with 17. Add 25. Then multiply the result by 8. What is the final value? Show all steps.,"Step 1: 17 + 25 = 42
Step 2: 42 * 8 = 336
Final answer: 336","math,arithmetic,step-by-step"
arb_1675,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1676,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1677,mathematical_reasoning,arithmetic,easy,Start with 28. Subtract 40. Then add 7. What is the final value? Show all steps.,"Step 1: 28 - 40 = -12
Step 2: -12 + 7 = -5
Final answer: -5","math,arithmetic,step-by-step"
arb_1678,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the umbrella from the warehouse and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the office.
6. Deliver / place the umbrella in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1679,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1680,logical_reasoning,deduction,medium,"Either Jamie or Riley took the key, but not both. Jamie did not take the key. Who took it?",Riley took the key. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1681,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1682,mathematical_reasoning,arithmetic,easy,Compute (42 * 31) + 30. Explain the order of operations.,"Multiplication first: 42 * 31 = 1302
Then addition: 1302 + 30 = 1332
Final answer: 1332","math,arithmetic,step-by-step"
arb_1683,tool_use,selection,easy,"Available tools: terminal, calendar, database_query, code_interpreter, browser.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1684,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_1685,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1686,logical_reasoning,deduction,easy,No Quinns are in the kitchen. Morgan is in the kitchen. Is Morgan a Quinn? Explain.,"No. The premise states that no Quinns are in the kitchen. Since Morgan is in the kitchen, Morgan cannot be a Quinn.","deduction,syllogism,logic"
arb_1687,commonsense_reasoning,everyday,easy,A green cat is observed in the garage. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1688,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1689,logical_reasoning,deduction,medium,No Taylors are in the garage. Jordan is in the garage. Is Jordan a Taylor? Explain.,"No. The premise states that no Taylors are in the garage. Since Jordan is in the garage, Jordan cannot be a Taylor.","deduction,syllogism,logic"
arb_1690,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the umbrella from the meeting room and deliver it to the warehouse. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the umbrella (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the warehouse.
6. Deliver / place the umbrella in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1691,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1692,mathematical_reasoning,arithmetic,easy,Start with 17. Subtract 44. Then add 36. What is the final value? Show all steps.,"Step 1: 17 - 44 = -27
Step 2: -27 + 36 = 9
Final answer: 9","math,arithmetic,step-by-step"
arb_1693,mathematical_reasoning,arithmetic,easy,Compute (30 * 49) + 29. Explain the order of operations.,"Multiplication first: 30 * 49 = 1470
Then addition: 1470 + 29 = 1499
Final answer: 1499","math,arithmetic,step-by-step"
arb_1694,logical_reasoning,deduction,medium,No Taylors are in the meeting room. Alex is in the meeting room. Is Alex a Taylor? Explain.,"No. The premise states that no Taylors are in the meeting room. Since Alex is in the meeting room, Alex cannot be a Taylor.","deduction,syllogism,logic"
arb_1695,mathematical_reasoning,arithmetic,easy,Start with 29. Add 15. Then multiply the result by 33. What is the final value? Show all steps.,"Step 1: 29 + 15 = 44
Step 2: 44 * 33 = 1452
Final answer: 1452","math,arithmetic,step-by-step"
arb_1696,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1697,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the key from the bedroom and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the server room.
6. Deliver / place the key in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1698,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_1699,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1700,mathematical_reasoning,arithmetic,easy,Compute (16 * 24) + 38. Explain the order of operations.,"Multiplication first: 16 * 24 = 384
Then addition: 384 + 38 = 422
Final answer: 422","math,arithmetic,step-by-step"
arb_1701,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1702,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1703,logical_reasoning,deduction,easy,No Rileys are in the lobby. Jamie is in the lobby. Is Jamie a Riley? Explain.,"No. The premise states that no Rileys are in the lobby. Since Jamie is in the lobby, Jamie cannot be a Riley.","deduction,syllogism,logic"
arb_1704,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_1705,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the wallet from the library and deliver it to the garage. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1706,commonsense_reasoning,everyday,easy,A blue fish is observed in the meeting room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1707,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the charger from the garage and deliver it to the lab. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the charger (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1708,mathematical_reasoning,arithmetic,easy,Start with 5. Subtract 9. Then add 4. What is the final value? Show all steps.,"Step 1: 5 - 9 = -4
Step 2: -4 + 4 = 0
Final answer: 0","math,arithmetic,step-by-step"
arb_1709,commonsense_reasoning,everyday,easy,A purple fish is observed in the meeting room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1710,tool_use,selection,easy,"Available tools: web_search, browser, database_query, email_client, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1711,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1712,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1713,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1714,mathematical_reasoning,arithmetic,easy,Start with 24. Subtract 40. Then add 12. What is the final value? Show all steps.,"Step 1: 24 - 40 = -16
Step 2: -16 + 12 = -4
Final answer: -4","math,arithmetic,step-by-step"
arb_1715,mathematical_reasoning,arithmetic,easy,Start with 10. Subtract 48. Then add 45. What is the final value? Show all steps.,"Step 1: 10 - 48 = -38
Step 2: -38 + 45 = 7
Final answer: 7","math,arithmetic,step-by-step"
arb_1716,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1717,logical_reasoning,deduction,easy,"Either Jamie or Cameron took the badge, but not both. Jamie did not take the badge. Who took it?",Cameron took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1718,tool_use,selection,easy,"Available tools: database_query, terminal, file_system, email_client.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1719,tool_use,selection,medium,"Available tools: code_interpreter, browser, email_client.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1720,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_1721,logical_reasoning,deduction,medium,No Taylors are in the bedroom. Quinn is in the bedroom. Is Quinn a Taylor? Explain.,"No. The premise states that no Taylors are in the bedroom. Since Quinn is in the bedroom, Quinn cannot be a Taylor.","deduction,syllogism,logic"
arb_1722,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the key from the garage and deliver it to the lab. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lab.
6. Deliver / place the key in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1723,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1724,commonsense_reasoning,everyday,easy,A purple rabbit is observed in the server room. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1725,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1726,mathematical_reasoning,arithmetic,easy,Compute (22 * 35) + 3. Explain the order of operations.,"Multiplication first: 22 * 35 = 770
Then addition: 770 + 3 = 773
Final answer: 773","math,arithmetic,step-by-step"
arb_1727,mathematical_reasoning,arithmetic,easy,Compute (20 * 6) + 3. Explain the order of operations.,"Multiplication first: 20 * 6 = 120
Then addition: 120 + 3 = 123
Final answer: 123","math,arithmetic,step-by-step"
arb_1728,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1729,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the service downtime.","causal,diagnosis,agentic"
arb_1730,tool_use,selection,easy,"Available tools: web_search, database_query, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1731,logical_reasoning,deduction,easy,"If someone owns a notebook, then they have access to the kitchen. Riley owns a notebook. What can we conclude about Riley?","We can conclude that Riley has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1732,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_1733,logical_reasoning,deduction,easy,"Either Avery or Alex took the wallet, but not both. Avery did not take the wallet. Who took it?",Alex took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1734,commonsense_reasoning,everyday,easy,A green bird is observed in the lab. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1735,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the wallet from the office and deliver it to the lab. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the wallet (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the lab.
6. Deliver / place the wallet in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1736,mathematical_reasoning,arithmetic,easy,Compute (8 * 29) + 20. Explain the order of operations.,"Multiplication first: 8 * 29 = 232
Then addition: 232 + 20 = 252
Final answer: 252","math,arithmetic,step-by-step"
arb_1737,commonsense_reasoning,everyday,easy,A blue bird is observed in the bedroom. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1738,commonsense_reasoning,everyday,easy,A purple dog is observed in the library. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1739,tool_use,selection,medium,"Available tools: file_system, calendar, database_query, terminal.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1740,commonsense_reasoning,everyday,easy,A black dog is observed in the library. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1741,tool_use,selection,easy,"Available tools: calendar, terminal, email_client, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1742,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_1743,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the wallet from the library and deliver it to the warehouse. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the warehouse.
6. Deliver / place the wallet in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1744,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1745,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1746,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the service downtime.","causal,diagnosis,agentic"
arb_1747,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1748,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1749,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_1750,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1751,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the error logs.","causal,diagnosis,agentic"
arb_1752,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the pen from the office and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1753,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the kitchen. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1754,mathematical_reasoning,arithmetic,easy,Compute (4 * 20) + 40. Explain the order of operations.,"Multiplication first: 4 * 20 = 80
Then addition: 80 + 40 = 120
Final answer: 120","math,arithmetic,step-by-step"
arb_1755,tool_use,selection,medium,"Available tools: terminal, code_interpreter, calendar, file_system, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1756,commonsense_reasoning,everyday,easy,A green dog is observed in the kitchen. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1757,mathematical_reasoning,arithmetic,easy,Compute (36 * 13) + 43. Explain the order of operations.,"Multiplication first: 36 * 13 = 468
Then addition: 468 + 43 = 511
Final answer: 511","math,arithmetic,step-by-step"
arb_1758,logical_reasoning,deduction,easy,All Averys own a umbrella. Alex is a Avery. Does Alex own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Averys own a umbrella' and the particular 'Alex is a Avery', it follows by universal instantiation that Alex owns a umbrella.","deduction,syllogism,logic"
arb_1759,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the laptop from the server room and deliver it to the warehouse. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the laptop (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1760,logical_reasoning,deduction,medium,"Either Quinn or Morgan took the wallet, but not both. Quinn did not take the wallet. Who took it?",Morgan took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1761,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_1762,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1763,logical_reasoning,deduction,medium,All Quinns own a badge. Jordan is a Quinn. Does Jordan own a badge? Explain step by step.,"Yes. From the universal statement 'All Quinns own a badge' and the particular 'Jordan is a Quinn', it follows by universal instantiation that Jordan owns a badge.","deduction,syllogism,logic"
arb_1764,logical_reasoning,deduction,easy,"Either Riley or Alex took the umbrella, but not both. Riley did not take the umbrella. Who took it?",Alex took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1765,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1766,tool_use,selection,medium,"Available tools: terminal, file_system, browser, code_interpreter, web_search.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1767,commonsense_reasoning,everyday,easy,A green dog is observed in the lobby. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1768,mathematical_reasoning,arithmetic,easy,Start with 6. Add 4. Then multiply the result by 28. What is the final value? Show all steps.,"Step 1: 6 + 4 = 10
Step 2: 10 * 28 = 280
Final answer: 280","math,arithmetic,step-by-step"
arb_1769,commonsense_reasoning,everyday,easy,A white fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1770,tool_use,selection,medium,"Available tools: terminal, email_client, web_search, calendar.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1771,mathematical_reasoning,arithmetic,easy,Compute (42 * 23) + 29. Explain the order of operations.,"Multiplication first: 42 * 23 = 966
Then addition: 966 + 29 = 995
Final answer: 995","math,arithmetic,step-by-step"
arb_1772,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1773,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the umbrella from the warehouse and deliver it to the office. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the office.
6. Deliver / place the umbrella in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1774,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_1775,logical_reasoning,deduction,medium,All Taylors own a notebook. Casey is a Taylor. Does Casey own a notebook? Explain step by step.,"Yes. From the universal statement 'All Taylors own a notebook' and the particular 'Casey is a Taylor', it follows by universal instantiation that Casey owns a notebook.","deduction,syllogism,logic"
arb_1776,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the umbrella from the lab and deliver it to the lobby. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the umbrella (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the lobby.
6. Deliver / place the umbrella in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1777,mathematical_reasoning,arithmetic,easy,Compute (13 * 18) + 19. Explain the order of operations.,"Multiplication first: 13 * 18 = 234
Then addition: 234 + 19 = 253
Final answer: 253","math,arithmetic,step-by-step"
arb_1778,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1779,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_1780,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1781,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1782,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1783,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the key from the lobby and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the server room.
6. Deliver / place the key in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1784,mathematical_reasoning,arithmetic,easy,Start with 43. Add 29. Then multiply the result by 44. What is the final value? Show all steps.,"Step 1: 43 + 29 = 72
Step 2: 72 * 44 = 3168
Final answer: 3168","math,arithmetic,step-by-step"
arb_1785,mathematical_reasoning,arithmetic,easy,Compute (40 * 42) + 29. Explain the order of operations.,"Multiplication first: 40 * 42 = 1680
Then addition: 1680 + 29 = 1709
Final answer: 1709","math,arithmetic,step-by-step"
arb_1786,tool_use,selection,easy,"Available tools: database_query, terminal, calendar.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1787,tool_use,selection,medium,"Available tools: browser, database_query, web_search.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1788,commonsense_reasoning,everyday,easy,A orange cat is observed in the lab. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1789,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1790,commonsense_reasoning,everyday,easy,A black wolf is observed in the bedroom. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1791,logical_reasoning,deduction,medium,"Either Cameron or Taylor took the pen, but not both. Cameron did not take the pen. Who took it?",Taylor took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1792,tool_use,selection,medium,"Available tools: web_search, browser, code_interpreter, terminal, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1793,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1794,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1795,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_1796,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1797,commonsense_reasoning,everyday,easy,A orange dog is observed in the meeting room. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1798,logical_reasoning,deduction,easy,All Alexs own a umbrella. Jordan is a Alex. Does Jordan own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Alexs own a umbrella' and the particular 'Jordan is a Alex', it follows by universal instantiation that Jordan owns a umbrella.","deduction,syllogism,logic"
arb_1799,commonsense_reasoning,everyday,easy,A purple wolf is observed in the office. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1800,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1801,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_1802,tool_use,selection,easy,"Available tools: file_system, database_query, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1803,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the key from the server room and deliver it to the kitchen. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the kitchen.
6. Deliver / place the key in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1804,mathematical_reasoning,arithmetic,easy,Compute (33 * 2) + 8. Explain the order of operations.,"Multiplication first: 33 * 2 = 66
Then addition: 66 + 8 = 74
Final answer: 74","math,arithmetic,step-by-step"
arb_1805,commonsense_reasoning,everyday,easy,A purple fish is observed in the garage. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1806,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1807,tool_use,selection,easy,"Available tools: web_search, browser, calendar, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1808,logical_reasoning,deduction,medium,"Either Alex or Cameron took the pen, but not both. Alex did not take the pen. Who took it?",Cameron took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1809,mathematical_reasoning,arithmetic,easy,Compute (2 * 45) + 9. Explain the order of operations.,"Multiplication first: 2 * 45 = 90
Then addition: 90 + 9 = 99
Final answer: 99","math,arithmetic,step-by-step"
arb_1810,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the phone from the lab and deliver it to the bedroom. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the phone (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the bedroom.
6. Deliver / place the phone in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1811,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1812,mathematical_reasoning,arithmetic,easy,Start with 31. Subtract 48. Then add 27. What is the final value? Show all steps.,"Step 1: 31 - 48 = -17
Step 2: -17 + 27 = 10
Final answer: 10","math,arithmetic,step-by-step"
arb_1813,tool_use,selection,medium,"Available tools: calendar, email_client, database_query, file_system.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1814,mathematical_reasoning,arithmetic,easy,Start with 8. Add 15. Then multiply the result by 46. What is the final value? Show all steps.,"Step 1: 8 + 15 = 23
Step 2: 23 * 46 = 1058
Final answer: 1058","math,arithmetic,step-by-step"
arb_1815,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1816,tool_use,selection,medium,"Available tools: code_interpreter, terminal, calendar, email_client, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1817,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the charger from the lab and deliver it to the office. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the charger (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the office.
6. Deliver / place the charger in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1818,tool_use,selection,medium,"Available tools: database_query, code_interpreter, browser, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1819,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1820,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1821,tool_use,selection,medium,"Available tools: code_interpreter, calendar, file_system, terminal, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1822,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1823,logical_reasoning,deduction,easy,No Alexs are in the lobby. Jamie is in the lobby. Is Jamie a Alex? Explain.,"No. The premise states that no Alexs are in the lobby. Since Jamie is in the lobby, Jamie cannot be a Alex.","deduction,syllogism,logic"
arb_1824,logical_reasoning,deduction,easy,No Jamies are in the meeting room. Casey is in the meeting room. Is Casey a Jamie? Explain.,"No. The premise states that no Jamies are in the meeting room. Since Casey is in the meeting room, Casey cannot be a Jamie.","deduction,syllogism,logic"
arb_1825,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1826,tool_use,selection,medium,"Available tools: terminal, file_system, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1827,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1828,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1829,tool_use,selection,medium,"Available tools: email_client, web_search, browser, calendar.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1830,logical_reasoning,deduction,easy,No Taylors are in the kitchen. Morgan is in the kitchen. Is Morgan a Taylor? Explain.,"No. The premise states that no Taylors are in the kitchen. Since Morgan is in the kitchen, Morgan cannot be a Taylor.","deduction,syllogism,logic"
arb_1831,tool_use,selection,medium,"Available tools: browser, database_query, terminal, email_client.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1832,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1833,tool_use,selection,easy,"Available tools: database_query, email_client, code_interpreter, terminal, web_search.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1834,mathematical_reasoning,arithmetic,easy,Start with 28. Subtract 29. Then add 45. What is the final value? Show all steps.,"Step 1: 28 - 29 = -1
Step 2: -1 + 45 = 44
Final answer: 44","math,arithmetic,step-by-step"
arb_1835,mathematical_reasoning,arithmetic,easy,Start with 1. Subtract 44. Then add 12. What is the final value? Show all steps.,"Step 1: 1 - 44 = -43
Step 2: -43 + 12 = -31
Final answer: -31","math,arithmetic,step-by-step"
arb_1836,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1837,mathematical_reasoning,arithmetic,easy,Compute (44 * 25) + 23. Explain the order of operations.,"Multiplication first: 44 * 25 = 1100
Then addition: 1100 + 23 = 1123
Final answer: 1123","math,arithmetic,step-by-step"
arb_1838,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1839,logical_reasoning,deduction,medium,No Alexs are in the lobby. Morgan is in the lobby. Is Morgan a Alex? Explain.,"No. The premise states that no Alexs are in the lobby. Since Morgan is in the lobby, Morgan cannot be a Alex.","deduction,syllogism,logic"
arb_1840,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the phone from the lobby and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1841,commonsense_reasoning,everyday,easy,A blue fish is observed in the library. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1842,logical_reasoning,deduction,easy,"Either Jamie or Taylor took the charger, but not both. Jamie did not take the charger. Who took it?",Taylor took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1843,tool_use,selection,medium,"Available tools: file_system, web_search, terminal, database_query.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1844,mathematical_reasoning,arithmetic,easy,Start with 26. Subtract 3. Then add 6. What is the final value? Show all steps.,"Step 1: 26 - 3 = 23
Step 2: 23 + 6 = 29
Final answer: 29","math,arithmetic,step-by-step"
arb_1845,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1846,tool_use,selection,easy,"Available tools: code_interpreter, terminal, email_client, calendar, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1847,commonsense_reasoning,everyday,easy,A black fish is observed in the library. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1848,logical_reasoning,deduction,medium,"Either Casey or Taylor took the book, but not both. Casey did not take the book. Who took it?",Taylor took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1849,logical_reasoning,deduction,medium,All Morgans own a charger. Avery is a Morgan. Does Avery own a charger? Explain step by step.,"Yes. From the universal statement 'All Morgans own a charger' and the particular 'Avery is a Morgan', it follows by universal instantiation that Avery owns a charger.","deduction,syllogism,logic"
arb_1850,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1851,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1852,tool_use,selection,easy,"Available tools: database_query, terminal, file_system.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1853,tool_use,selection,easy,"Available tools: file_system, email_client, database_query, terminal.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1854,commonsense_reasoning,everyday,easy,A red cat is observed in the garage. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1855,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1856,commonsense_reasoning,everyday,easy,A blue bird is observed in the warehouse. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1857,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1858,tool_use,selection,easy,"Available tools: email_client, terminal, file_system.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1859,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the charger from the server room and deliver it to the kitchen. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the charger (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the kitchen.
6. Deliver / place the charger in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1860,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1861,tool_use,selection,medium,"Available tools: calendar, code_interpreter, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1862,tool_use,selection,medium,"Available tools: browser, email_client, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1863,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1864,commonsense_reasoning,everyday,easy,A orange cat is observed in the kitchen. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1865,mathematical_reasoning,arithmetic,easy,Start with 18. Add 13. Then multiply the result by 44. What is the final value? Show all steps.,"Step 1: 18 + 13 = 31
Step 2: 31 * 44 = 1364
Final answer: 1364","math,arithmetic,step-by-step"
arb_1866,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the wallet from the server room and deliver it to the office. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the wallet (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the office.
6. Deliver / place the wallet in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1867,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the book from the warehouse and deliver it to the meeting room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the book (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the meeting room.
6. Deliver / place the book in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1868,tool_use,selection,easy,"Available tools: calendar, web_search, terminal, browser, code_interpreter.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1869,tool_use,selection,easy,"Available tools: calendar, file_system, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1870,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the notebook from the garage and deliver it to the library. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the notebook (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the library.
6. Deliver / place the notebook in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1871,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1872,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1873,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_1874,tool_use,selection,easy,"Available tools: web_search, email_client, file_system, terminal.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1875,logical_reasoning,deduction,medium,"Either Jordan or Casey took the notebook, but not both. Jordan did not take the notebook. Who took it?",Casey took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1876,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the notebook from the warehouse and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the notebook (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the server room.
6. Deliver / place the notebook in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1877,mathematical_reasoning,arithmetic,easy,Start with 36. Subtract 45. Then add 46. What is the final value? Show all steps.,"Step 1: 36 - 45 = -9
Step 2: -9 + 46 = 37
Final answer: 37","math,arithmetic,step-by-step"
arb_1878,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1879,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1880,logical_reasoning,deduction,medium,All Caseys own a wallet. Alex is a Casey. Does Alex own a wallet? Explain step by step.,"Yes. From the universal statement 'All Caseys own a wallet' and the particular 'Alex is a Casey', it follows by universal instantiation that Alex owns a wallet.","deduction,syllogism,logic"
arb_1881,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the wallet from the lobby and deliver it to the office. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the wallet (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the office.
6. Deliver / place the wallet in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1882,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1883,mathematical_reasoning,arithmetic,easy,Compute (14 * 16) + 46. Explain the order of operations.,"Multiplication first: 14 * 16 = 224
Then addition: 224 + 46 = 270
Final answer: 270","math,arithmetic,step-by-step"
arb_1884,logical_reasoning,deduction,medium,"If someone owns a key, then they have access to the lobby. Taylor owns a key. What can we conclude about Taylor?","We can conclude that Taylor has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_1885,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1886,logical_reasoning,deduction,easy,All Caseys own a key. Morgan is a Casey. Does Morgan own a key? Explain step by step.,"Yes. From the universal statement 'All Caseys own a key' and the particular 'Morgan is a Casey', it follows by universal instantiation that Morgan owns a key.","deduction,syllogism,logic"
arb_1887,mathematical_reasoning,arithmetic,easy,Start with 46. Add 28. Then multiply the result by 27. What is the final value? Show all steps.,"Step 1: 46 + 28 = 74
Step 2: 74 * 27 = 1998
Final answer: 1998","math,arithmetic,step-by-step"
arb_1888,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1889,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1890,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1891,mathematical_reasoning,arithmetic,easy,Compute (15 * 39) + 43. Explain the order of operations.,"Multiplication first: 15 * 39 = 585
Then addition: 585 + 43 = 628
Final answer: 628","math,arithmetic,step-by-step"
arb_1892,mathematical_reasoning,arithmetic,easy,Start with 48. Subtract 1. Then add 30. What is the final value? Show all steps.,"Step 1: 48 - 1 = 47
Step 2: 47 + 30 = 77
Final answer: 77","math,arithmetic,step-by-step"
arb_1893,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1894,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the laptop from the kitchen and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the meeting room.
6. Deliver / place the laptop in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1895,mathematical_reasoning,arithmetic,easy,Start with 24. Subtract 38. Then add 7. What is the final value? Show all steps.,"Step 1: 24 - 38 = -14
Step 2: -14 + 7 = -7
Final answer: -7","math,arithmetic,step-by-step"
arb_1896,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1897,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1898,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the laptop from the kitchen and deliver it to the warehouse. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1899,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1900,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1901,tool_use,selection,easy,"Available tools: email_client, calendar, browser.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1902,tool_use,selection,medium,"Available tools: email_client, code_interpreter, browser, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1903,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the charger from the lab and deliver it to the office. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the charger (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the office.
6. Deliver / place the charger in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1904,logical_reasoning,deduction,easy,No Jamies are in the meeting room. Taylor is in the meeting room. Is Taylor a Jamie? Explain.,"No. The premise states that no Jamies are in the meeting room. Since Taylor is in the meeting room, Taylor cannot be a Jamie.","deduction,syllogism,logic"
arb_1905,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1906,logical_reasoning,deduction,medium,"Either Jordan or Avery took the notebook, but not both. Jordan did not take the notebook. Who took it?",Avery took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1907,tool_use,selection,medium,"Available tools: database_query, calendar, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1908,tool_use,selection,medium,"Available tools: code_interpreter, calendar, web_search.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1909,tool_use,selection,easy,"Available tools: file_system, database_query, code_interpreter.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1910,commonsense_reasoning,everyday,easy,A yellow fish is observed in the kitchen. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1911,logical_reasoning,deduction,easy,"If someone owns a umbrella, then they have access to the kitchen. Alex owns a umbrella. What can we conclude about Alex?","We can conclude that Alex has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1912,tool_use,selection,easy,"Available tools: calendar, email_client, code_interpreter, browser, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1913,logical_reasoning,deduction,easy,"Either Cameron or Avery took the wallet, but not both. Cameron did not take the wallet. Who took it?",Avery took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1914,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1915,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the notebook from the bedroom and deliver it to the kitchen. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the notebook (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the kitchen.
6. Deliver / place the notebook in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1916,logical_reasoning,deduction,medium,All Camerons own a pen. Jamie is a Cameron. Does Jamie own a pen? Explain step by step.,"Yes. From the universal statement 'All Camerons own a pen' and the particular 'Jamie is a Cameron', it follows by universal instantiation that Jamie owns a pen.","deduction,syllogism,logic"
arb_1917,mathematical_reasoning,arithmetic,easy,Start with 44. Subtract 38. Then add 21. What is the final value? Show all steps.,"Step 1: 44 - 38 = 6
Step 2: 6 + 21 = 27
Final answer: 27","math,arithmetic,step-by-step"
arb_1918,logical_reasoning,deduction,medium,"Either Jordan or Casey took the pen, but not both. Jordan did not take the pen. Who took it?",Casey took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1919,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1920,mathematical_reasoning,arithmetic,easy,Start with 47. Subtract 48. Then add 21. What is the final value? Show all steps.,"Step 1: 47 - 48 = -1
Step 2: -1 + 21 = 20
Final answer: 20","math,arithmetic,step-by-step"
arb_1921,logical_reasoning,deduction,easy,All Jordans own a pen. Cameron is a Jordan. Does Cameron own a pen? Explain step by step.,"Yes. From the universal statement 'All Jordans own a pen' and the particular 'Cameron is a Jordan', it follows by universal instantiation that Cameron owns a pen.","deduction,syllogism,logic"
arb_1922,logical_reasoning,deduction,easy,No Jordans are in the garage. Casey is in the garage. Is Casey a Jordan? Explain.,"No. The premise states that no Jordans are in the garage. Since Casey is in the garage, Casey cannot be a Jordan.","deduction,syllogism,logic"
arb_1923,logical_reasoning,deduction,easy,"Either Riley or Quinn took the phone, but not both. Riley did not take the phone. Who took it?",Quinn took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1924,logical_reasoning,deduction,medium,"Either Jordan or Cameron took the wallet, but not both. Jordan did not take the wallet. Who took it?",Cameron took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1925,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the badge from the library and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the badge (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the office.
6. Deliver / place the badge in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1926,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_1927,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1928,tool_use,selection,easy,"Available tools: code_interpreter, database_query, file_system, email_client, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1929,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1930,logical_reasoning,deduction,medium,"Either Jamie or Cameron took the badge, but not both. Jamie did not take the badge. Who took it?",Cameron took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1931,commonsense_reasoning,everyday,easy,A orange fish is observed in the bedroom. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1932,commonsense_reasoning,everyday,easy,A red fish is observed in the garage. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1933,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the charger from the warehouse and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the charger (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the server room.
6. Deliver / place the charger in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1934,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1935,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1936,logical_reasoning,deduction,medium,"Either Morgan or Casey took the charger, but not both. Morgan did not take the charger. Who took it?",Casey took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1937,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1938,mathematical_reasoning,arithmetic,easy,Start with 6. Subtract 10. Then add 20. What is the final value? Show all steps.,"Step 1: 6 - 10 = -4
Step 2: -4 + 20 = 16
Final answer: 16","math,arithmetic,step-by-step"
arb_1939,mathematical_reasoning,arithmetic,easy,Start with 41. Add 22. Then multiply the result by 17. What is the final value? Show all steps.,"Step 1: 41 + 22 = 63
Step 2: 63 * 17 = 1071
Final answer: 1071","math,arithmetic,step-by-step"
arb_1940,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1941,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the umbrella from the library and deliver it to the kitchen. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the umbrella (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the kitchen.
6. Deliver / place the umbrella in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1942,logical_reasoning,deduction,medium,"If someone owns a charger, then they have access to the lab. Taylor owns a charger. What can we conclude about Taylor?","We can conclude that Taylor has access to the lab, by modus ponens.","deduction,syllogism,logic"
arb_1943,tool_use,selection,easy,"Available tools: web_search, database_query, calendar, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1944,commonsense_reasoning,everyday,easy,A yellow rabbit is observed in the meeting room. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1945,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1946,logical_reasoning,deduction,medium,"If someone owns a phone, then they have access to the kitchen. Avery owns a phone. What can we conclude about Avery?","We can conclude that Avery has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1947,mathematical_reasoning,arithmetic,easy,Start with 42. Subtract 9. Then add 1. What is the final value? Show all steps.,"Step 1: 42 - 9 = 33
Step 2: 33 + 1 = 34
Final answer: 34","math,arithmetic,step-by-step"
arb_1948,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the badge from the server room and deliver it to the meeting room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the meeting room.
6. Deliver / place the badge in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1949,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1950,logical_reasoning,deduction,easy,No Taylors are in the office. Cameron is in the office. Is Cameron a Taylor? Explain.,"No. The premise states that no Taylors are in the office. Since Cameron is in the office, Cameron cannot be a Taylor.","deduction,syllogism,logic"
arb_1951,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_1952,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1953,tool_use,selection,easy,"Available tools: terminal, browser, database_query, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1954,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1955,mathematical_reasoning,arithmetic,easy,Compute (24 * 31) + 11. Explain the order of operations.,"Multiplication first: 24 * 31 = 744
Then addition: 744 + 11 = 755
Final answer: 755","math,arithmetic,step-by-step"
arb_1956,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_1957,logical_reasoning,deduction,medium,"Either Cameron or Jordan took the umbrella, but not both. Cameron did not take the umbrella. Who took it?",Jordan took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1958,logical_reasoning,deduction,medium,"If someone owns a key, then they have access to the lobby. Avery owns a key. What can we conclude about Avery?","We can conclude that Avery has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_1959,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1960,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1961,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_1962,tool_use,selection,easy,"Available tools: web_search, browser, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1963,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1964,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the key from the server room and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1965,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1966,logical_reasoning,deduction,easy,No Camerons are in the lobby. Riley is in the lobby. Is Riley a Cameron? Explain.,"No. The premise states that no Camerons are in the lobby. Since Riley is in the lobby, Riley cannot be a Cameron.","deduction,syllogism,logic"
arb_1967,mathematical_reasoning,arithmetic,easy,Compute (49 * 33) + 32. Explain the order of operations.,"Multiplication first: 49 * 33 = 1617
Then addition: 1617 + 32 = 1649
Final answer: 1649","math,arithmetic,step-by-step"
arb_1968,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1969,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the kitchen. Cameron owns a laptop. What can we conclude about Cameron?","We can conclude that Cameron has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_1970,mathematical_reasoning,arithmetic,easy,Start with 42. Add 11. Then multiply the result by 1. What is the final value? Show all steps.,"Step 1: 42 + 11 = 53
Step 2: 53 * 1 = 53
Final answer: 53","math,arithmetic,step-by-step"
arb_1971,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1972,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the phone from the warehouse and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the phone (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the meeting room.
6. Deliver / place the phone in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_1973,logical_reasoning,deduction,medium,"If someone owns a badge, then they have access to the meeting room. Taylor owns a badge. What can we conclude about Taylor?","We can conclude that Taylor has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_1974,logical_reasoning,deduction,medium,"If someone owns a book, then they have access to the bedroom. Riley owns a book. What can we conclude about Riley?","We can conclude that Riley has access to the bedroom, by modus ponens.","deduction,syllogism,logic"
arb_1975,logical_reasoning,deduction,easy,No Jordans are in the bedroom. Morgan is in the bedroom. Is Morgan a Jordan? Explain.,"No. The premise states that no Jordans are in the bedroom. Since Morgan is in the bedroom, Morgan cannot be a Jordan.","deduction,syllogism,logic"
arb_1976,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1977,tool_use,selection,easy,"Available tools: email_client, calendar, terminal.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1978,mathematical_reasoning,arithmetic,easy,Compute (16 * 34) + 46. Explain the order of operations.,"Multiplication first: 16 * 34 = 544
Then addition: 544 + 46 = 590
Final answer: 590","math,arithmetic,step-by-step"
arb_1979,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1980,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_1981,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1982,tool_use,selection,medium,"Available tools: terminal, file_system, web_search, code_interpreter.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1983,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1984,mathematical_reasoning,arithmetic,easy,Compute (27 * 18) + 8. Explain the order of operations.,"Multiplication first: 27 * 18 = 486
Then addition: 486 + 8 = 494
Final answer: 494","math,arithmetic,step-by-step"
arb_1985,tool_use,selection,easy,"Available tools: web_search, database_query, email_client, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1986,logical_reasoning,deduction,easy,"If someone owns a pen, then they have access to the garage. Quinn owns a pen. What can we conclude about Quinn?","We can conclude that Quinn has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_1987,tool_use,selection,medium,"Available tools: email_client, terminal, file_system, browser, web_search.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1988,commonsense_reasoning,everyday,easy,A red cat is observed in the meeting room. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_1989,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_1990,logical_reasoning,deduction,medium,No Quinns are in the kitchen. Jamie is in the kitchen. Is Jamie a Quinn? Explain.,"No. The premise states that no Quinns are in the kitchen. Since Jamie is in the kitchen, Jamie cannot be a Quinn.","deduction,syllogism,logic"
arb_1991,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_1992,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1993,mathematical_reasoning,arithmetic,easy,Start with 29. Subtract 40. Then add 17. What is the final value? Show all steps.,"Step 1: 29 - 40 = -11
Step 2: -11 + 17 = 6
Final answer: 6","math,arithmetic,step-by-step"
arb_1994,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_1995,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_1996,mathematical_reasoning,arithmetic,easy,Compute (33 * 9) + 32. Explain the order of operations.,"Multiplication first: 33 * 9 = 297
Then addition: 297 + 32 = 329
Final answer: 329","math,arithmetic,step-by-step"
arb_1997,logical_reasoning,deduction,easy,"Either Taylor or Jordan took the wallet, but not both. Taylor did not take the wallet. Who took it?",Jordan took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_1998,tool_use,selection,easy,"Available tools: file_system, email_client, code_interpreter, calendar.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_1999,tool_use,selection,medium,"Available tools: email_client, browser, code_interpreter, database_query.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2000,mathematical_reasoning,arithmetic,easy,Compute (31 * 8) + 32. Explain the order of operations.,"Multiplication first: 31 * 8 = 248
Then addition: 248 + 32 = 280
Final answer: 280","math,arithmetic,step-by-step"
arb_2001,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the badge from the warehouse and deliver it to the library. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the badge (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the library.
6. Deliver / place the badge in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2002,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2003,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the phone from the meeting room and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the phone (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the office.
6. Deliver / place the phone in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2004,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2005,mathematical_reasoning,arithmetic,easy,Start with 44. Subtract 9. Then add 46. What is the final value? Show all steps.,"Step 1: 44 - 9 = 35
Step 2: 35 + 46 = 81
Final answer: 81","math,arithmetic,step-by-step"
arb_2006,logical_reasoning,deduction,easy,"Either Alex or Jamie took the notebook, but not both. Alex did not take the notebook. Who took it?",Jamie took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2007,commonsense_reasoning,everyday,easy,A orange bird is observed in the library. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a library (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2008,logical_reasoning,deduction,medium,"Either Casey or Avery took the pen, but not both. Casey did not take the pen. Who took it?",Avery took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2009,tool_use,selection,easy,"Available tools: calendar, browser, terminal.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2010,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_2011,commonsense_reasoning,everyday,easy,A black fox is observed in the lobby. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2012,logical_reasoning,deduction,easy,No Morgans are in the bedroom. Avery is in the bedroom. Is Avery a Morgan? Explain.,"No. The premise states that no Morgans are in the bedroom. Since Avery is in the bedroom, Avery cannot be a Morgan.","deduction,syllogism,logic"
arb_2013,mathematical_reasoning,arithmetic,easy,Start with 42. Subtract 34. Then add 6. What is the final value? Show all steps.,"Step 1: 42 - 34 = 8
Step 2: 8 + 6 = 14
Final answer: 14","math,arithmetic,step-by-step"
arb_2014,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2015,logical_reasoning,deduction,easy,All Rileys own a charger. Jamie is a Riley. Does Jamie own a charger? Explain step by step.,"Yes. From the universal statement 'All Rileys own a charger' and the particular 'Jamie is a Riley', it follows by universal instantiation that Jamie owns a charger.","deduction,syllogism,logic"
arb_2016,commonsense_reasoning,everyday,easy,A orange wolf is observed in the server room. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2017,tool_use,selection,easy,"Available tools: calendar, email_client, database_query, terminal, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2018,commonsense_reasoning,everyday,easy,A orange owl is observed in the server room. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a server room (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2019,tool_use,selection,medium,"Available tools: code_interpreter, calendar, database_query, browser, terminal.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2020,tool_use,selection,easy,"Available tools: code_interpreter, terminal, email_client, web_search, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2021,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2022,tool_use,selection,medium,"Available tools: file_system, calendar, browser, email_client, database_query.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2023,tool_use,selection,medium,"Available tools: browser, code_interpreter, email_client, file_system, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2024,tool_use,selection,easy,"Available tools: calendar, browser, file_system.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2025,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the error logs.","causal,diagnosis,agentic"
arb_2026,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the error logs.","causal,diagnosis,agentic"
arb_2027,commonsense_reasoning,everyday,easy,A yellow fox is observed in the office. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2028,tool_use,selection,medium,"Available tools: terminal, browser, web_search, code_interpreter.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2029,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the laptop from the garage and deliver it to the lab. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the laptop (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2030,commonsense_reasoning,everyday,easy,A black fox is observed in the warehouse. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2031,tool_use,selection,easy,"Available tools: database_query, terminal, code_interpreter, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2032,mathematical_reasoning,arithmetic,easy,Compute (14 * 39) + 19. Explain the order of operations.,"Multiplication first: 14 * 39 = 546
Then addition: 546 + 19 = 565
Final answer: 565","math,arithmetic,step-by-step"
arb_2033,logical_reasoning,deduction,medium,All Rileys own a book. Taylor is a Riley. Does Taylor own a book? Explain step by step.,"Yes. From the universal statement 'All Rileys own a book' and the particular 'Taylor is a Riley', it follows by universal instantiation that Taylor owns a book.","deduction,syllogism,logic"
arb_2034,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2035,tool_use,selection,medium,"Available tools: code_interpreter, database_query, terminal, calendar.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2036,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2037,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2038,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the user complaints.","causal,diagnosis,agentic"
arb_2039,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2040,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the notebook from the library and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the notebook (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2041,mathematical_reasoning,arithmetic,easy,Start with 44. Add 7. Then multiply the result by 22. What is the final value? Show all steps.,"Step 1: 44 + 7 = 51
Step 2: 51 * 22 = 1122
Final answer: 1122","math,arithmetic,step-by-step"
arb_2042,tool_use,selection,medium,"Available tools: email_client, web_search, browser, code_interpreter.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2043,mathematical_reasoning,arithmetic,easy,Compute (25 * 31) + 47. Explain the order of operations.,"Multiplication first: 25 * 31 = 775
Then addition: 775 + 47 = 822
Final answer: 822","math,arithmetic,step-by-step"
arb_2044,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2045,commonsense_reasoning,everyday,easy,A blue dog is observed in the garage. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2046,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the charger from the bedroom and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the server room.
6. Deliver / place the charger in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2047,logical_reasoning,deduction,medium,No Rileys are in the bedroom. Cameron is in the bedroom. Is Cameron a Riley? Explain.,"No. The premise states that no Rileys are in the bedroom. Since Cameron is in the bedroom, Cameron cannot be a Riley.","deduction,syllogism,logic"
arb_2048,mathematical_reasoning,arithmetic,easy,Start with 16. Subtract 20. Then add 47. What is the final value? Show all steps.,"Step 1: 16 - 20 = -4
Step 2: -4 + 47 = 43
Final answer: 43","math,arithmetic,step-by-step"
arb_2049,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2050,commonsense_reasoning,everyday,easy,A red fox is observed in the office. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2051,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2052,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the notebook from the kitchen and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the notebook (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the office.
6. Deliver / place the notebook in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2053,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2054,tool_use,selection,easy,"Available tools: code_interpreter, web_search, file_system, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2055,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the key from the garage and deliver it to the bedroom. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the key (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the bedroom.
6. Deliver / place the key in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2056,mathematical_reasoning,arithmetic,easy,Start with 36. Subtract 27. Then add 5. What is the final value? Show all steps.,"Step 1: 36 - 27 = 9
Step 2: 9 + 5 = 14
Final answer: 14","math,arithmetic,step-by-step"
arb_2057,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2058,tool_use,selection,medium,"Available tools: database_query, terminal, email_client.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2059,mathematical_reasoning,arithmetic,easy,Compute (30 * 9) + 46. Explain the order of operations.,"Multiplication first: 30 * 9 = 270
Then addition: 270 + 46 = 316
Final answer: 316","math,arithmetic,step-by-step"
arb_2060,mathematical_reasoning,arithmetic,easy,Compute (1 * 18) + 34. Explain the order of operations.,"Multiplication first: 1 * 18 = 18
Then addition: 18 + 34 = 52
Final answer: 52","math,arithmetic,step-by-step"
arb_2061,mathematical_reasoning,arithmetic,easy,Compute (11 * 4) + 26. Explain the order of operations.,"Multiplication first: 11 * 4 = 44
Then addition: 44 + 26 = 70
Final answer: 70","math,arithmetic,step-by-step"
arb_2062,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the pen from the kitchen and deliver it to the office. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the pen (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the office.
6. Deliver / place the pen in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2063,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_2064,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2065,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_2066,tool_use,selection,medium,"Available tools: code_interpreter, file_system, database_query, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2067,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_2068,commonsense_reasoning,everyday,easy,A blue fish is observed in the kitchen. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2069,logical_reasoning,deduction,medium,All Alexs own a laptop. Taylor is a Alex. Does Taylor own a laptop? Explain step by step.,"Yes. From the universal statement 'All Alexs own a laptop' and the particular 'Taylor is a Alex', it follows by universal instantiation that Taylor owns a laptop.","deduction,syllogism,logic"
arb_2070,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2071,mathematical_reasoning,arithmetic,easy,Start with 47. Subtract 13. Then add 31. What is the final value? Show all steps.,"Step 1: 47 - 13 = 34
Step 2: 34 + 31 = 65
Final answer: 65","math,arithmetic,step-by-step"
arb_2072,logical_reasoning,deduction,easy,"Either Jamie or Riley took the pen, but not both. Jamie did not take the pen. Who took it?",Riley took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2073,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2074,tool_use,selection,medium,"Available tools: terminal, browser, web_search, database_query, file_system.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2075,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the charger from the server room and deliver it to the lab. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the charger (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2076,mathematical_reasoning,arithmetic,easy,Start with 8. Subtract 32. Then add 3. What is the final value? Show all steps.,"Step 1: 8 - 32 = -24
Step 2: -24 + 3 = -21
Final answer: -21","math,arithmetic,step-by-step"
arb_2077,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the notebook from the server room and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the notebook (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2078,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the badge from the server room and deliver it to the warehouse. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the badge (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the warehouse.
6. Deliver / place the badge in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2079,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2080,mathematical_reasoning,arithmetic,easy,Start with 39. Add 17. Then multiply the result by 44. What is the final value? Show all steps.,"Step 1: 39 + 17 = 56
Step 2: 56 * 44 = 2464
Final answer: 2464","math,arithmetic,step-by-step"
arb_2081,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_2082,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Avery. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2083,tool_use,selection,easy,"Available tools: database_query, file_system, terminal, calendar, email_client.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2084,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2085,logical_reasoning,deduction,easy,"Either Morgan or Jamie took the charger, but not both. Morgan did not take the charger. Who took it?",Jamie took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2086,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2087,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2088,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2089,mathematical_reasoning,arithmetic,easy,Compute (23 * 8) + 4. Explain the order of operations.,"Multiplication first: 23 * 8 = 184
Then addition: 184 + 4 = 188
Final answer: 188","math,arithmetic,step-by-step"
arb_2090,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2091,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2092,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the umbrella from the meeting room and deliver it to the server room. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the server room.
6. Deliver / place the umbrella in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2093,mathematical_reasoning,arithmetic,easy,Compute (12 * 14) + 6. Explain the order of operations.,"Multiplication first: 12 * 14 = 168
Then addition: 168 + 6 = 174
Final answer: 174","math,arithmetic,step-by-step"
arb_2094,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2095,logical_reasoning,deduction,medium,"If someone owns a notebook, then they have access to the meeting room. Jamie owns a notebook. What can we conclude about Jamie?","We can conclude that Jamie has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_2096,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2097,logical_reasoning,deduction,medium,"Either Jamie or Alex took the wallet, but not both. Jamie did not take the wallet. Who took it?",Alex took the wallet. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2098,mathematical_reasoning,arithmetic,easy,Start with 8. Subtract 36. Then add 15. What is the final value? Show all steps.,"Step 1: 8 - 36 = -28
Step 2: -28 + 15 = -13
Final answer: -13","math,arithmetic,step-by-step"
arb_2099,logical_reasoning,deduction,medium,All Camerons own a charger. Avery is a Cameron. Does Avery own a charger? Explain step by step.,"Yes. From the universal statement 'All Camerons own a charger' and the particular 'Avery is a Cameron', it follows by universal instantiation that Avery owns a charger.","deduction,syllogism,logic"
arb_2100,mathematical_reasoning,arithmetic,easy,Compute (28 * 27) + 32. Explain the order of operations.,"Multiplication first: 28 * 27 = 756
Then addition: 756 + 32 = 788
Final answer: 788","math,arithmetic,step-by-step"
arb_2101,commonsense_reasoning,everyday,easy,A orange dog is observed in the library. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2102,mathematical_reasoning,arithmetic,easy,Compute (22 * 25) + 47. Explain the order of operations.,"Multiplication first: 22 * 25 = 550
Then addition: 550 + 47 = 597
Final answer: 597","math,arithmetic,step-by-step"
arb_2103,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the key from the bedroom and deliver it to the lobby. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the key (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lobby.
6. Deliver / place the key in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2104,mathematical_reasoning,arithmetic,easy,Compute (28 * 37) + 21. Explain the order of operations.,"Multiplication first: 28 * 37 = 1036
Then addition: 1036 + 21 = 1057
Final answer: 1057","math,arithmetic,step-by-step"
arb_2105,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the badge from the warehouse and deliver it to the lab. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the badge (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the lab.
6. Deliver / place the badge in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2106,commonsense_reasoning,everyday,easy,A purple fox is observed in the library. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2107,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the badge from the meeting room and deliver it to the bedroom. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the badge (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the bedroom.
6. Deliver / place the badge in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2108,commonsense_reasoning,everyday,easy,A yellow owl is observed in the office. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a office (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2109,logical_reasoning,deduction,medium,"Either Taylor or Quinn took the notebook, but not both. Taylor did not take the notebook. Who took it?",Quinn took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2110,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2111,mathematical_reasoning,arithmetic,easy,Compute (38 * 39) + 47. Explain the order of operations.,"Multiplication first: 38 * 39 = 1482
Then addition: 1482 + 47 = 1529
Final answer: 1529","math,arithmetic,step-by-step"
arb_2112,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_2113,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2114,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2115,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the wallet from the lab and deliver it to the garage. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the wallet (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2116,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_2117,logical_reasoning,deduction,medium,No Jamies are in the server room. Jordan is in the server room. Is Jordan a Jamie? Explain.,"No. The premise states that no Jamies are in the server room. Since Jordan is in the server room, Jordan cannot be a Jamie.","deduction,syllogism,logic"
arb_2118,tool_use,selection,easy,"Available tools: browser, file_system, email_client, web_search.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2119,mathematical_reasoning,arithmetic,easy,Start with 24. Subtract 44. Then add 34. What is the final value? Show all steps.,"Step 1: 24 - 44 = -20
Step 2: -20 + 34 = 14
Final answer: 14","math,arithmetic,step-by-step"
arb_2120,mathematical_reasoning,arithmetic,easy,Compute (25 * 26) + 19. Explain the order of operations.,"Multiplication first: 25 * 26 = 650
Then addition: 650 + 19 = 669
Final answer: 669","math,arithmetic,step-by-step"
arb_2121,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2122,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the user complaints.","causal,diagnosis,agentic"
arb_2123,mathematical_reasoning,arithmetic,easy,Compute (40 * 41) + 20. Explain the order of operations.,"Multiplication first: 40 * 41 = 1640
Then addition: 1640 + 20 = 1660
Final answer: 1660","math,arithmetic,step-by-step"
arb_2124,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2125,mathematical_reasoning,arithmetic,easy,Compute (25 * 46) + 28. Explain the order of operations.,"Multiplication first: 25 * 46 = 1150
Then addition: 1150 + 28 = 1178
Final answer: 1178","math,arithmetic,step-by-step"
arb_2126,tool_use,selection,easy,"Available tools: file_system, code_interpreter, web_search.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2127,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2128,mathematical_reasoning,arithmetic,easy,Start with 46. Subtract 43. Then add 31. What is the final value? Show all steps.,"Step 1: 46 - 43 = 3
Step 2: 3 + 31 = 34
Final answer: 34","math,arithmetic,step-by-step"
arb_2129,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the key from the office and deliver it to the garage. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the garage.
6. Deliver / place the key in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2130,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2131,tool_use,selection,easy,"Available tools: calendar, file_system, web_search.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2132,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the badge from the meeting room and deliver it to the lobby. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the badge (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the lobby.
6. Deliver / place the badge in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2133,tool_use,selection,easy,"Available tools: browser, email_client, database_query, terminal, code_interpreter.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2134,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_2135,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2136,tool_use,selection,easy,"Available tools: file_system, web_search, database_query, code_interpreter, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2137,logical_reasoning,deduction,medium,"Either Morgan or Jamie took the laptop, but not both. Morgan did not take the laptop. Who took it?",Jamie took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2138,commonsense_reasoning,everyday,easy,A green fox is observed in the garage. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2139,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_2140,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2141,tool_use,selection,easy,"Available tools: terminal, calendar, code_interpreter, email_client, database_query.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2142,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2143,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the charger from the bedroom and deliver it to the lobby. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lobby.
6. Deliver / place the charger in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2144,logical_reasoning,deduction,medium,"Either Taylor or Avery took the badge, but not both. Taylor did not take the badge. Who took it?",Avery took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2145,commonsense_reasoning,everyday,easy,A blue wolf is observed in the bedroom. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2146,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2147,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the badge from the meeting room and deliver it to the library. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the badge (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the library.
6. Deliver / place the badge in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2148,commonsense_reasoning,everyday,easy,A white fox is observed in the office. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2149,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2150,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2151,mathematical_reasoning,arithmetic,easy,Compute (4 * 41) + 46. Explain the order of operations.,"Multiplication first: 4 * 41 = 164
Then addition: 164 + 46 = 210
Final answer: 210","math,arithmetic,step-by-step"
arb_2152,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_2153,logical_reasoning,deduction,medium,All Alexs own a umbrella. Morgan is a Alex. Does Morgan own a umbrella? Explain step by step.,"Yes. From the universal statement 'All Alexs own a umbrella' and the particular 'Morgan is a Alex', it follows by universal instantiation that Morgan owns a umbrella.","deduction,syllogism,logic"
arb_2154,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2155,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2156,tool_use,selection,easy,"Available tools: code_interpreter, email_client, terminal, database_query.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2157,commonsense_reasoning,everyday,easy,A yellow wolf is observed in the warehouse. Is it more likely that the wolf lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about wolfs.,"Most wolfs do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the wolf is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2158,mathematical_reasoning,arithmetic,easy,Start with 36. Subtract 48. Then add 21. What is the final value? Show all steps.,"Step 1: 36 - 48 = -12
Step 2: -12 + 21 = 9
Final answer: 9","math,arithmetic,step-by-step"
arb_2159,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_2160,commonsense_reasoning,everyday,easy,A orange bird is observed in the meeting room. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2161,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2162,mathematical_reasoning,arithmetic,easy,Compute (34 * 49) + 3. Explain the order of operations.,"Multiplication first: 34 * 49 = 1666
Then addition: 1666 + 3 = 1669
Final answer: 1669","math,arithmetic,step-by-step"
arb_2163,mathematical_reasoning,arithmetic,easy,Start with 49. Add 41. Then multiply the result by 11. What is the final value? Show all steps.,"Step 1: 49 + 41 = 90
Step 2: 90 * 11 = 990
Final answer: 990","math,arithmetic,step-by-step"
arb_2164,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the badge from the office and deliver it to the server room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the badge (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the server room.
6. Deliver / place the badge in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2165,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the umbrella from the meeting room and deliver it to the kitchen. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the umbrella (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the kitchen.
6. Deliver / place the umbrella in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2166,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the umbrella from the garage and deliver it to the lobby. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the lobby.
6. Deliver / place the umbrella in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2167,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2168,commonsense_reasoning,everyday,easy,A orange rabbit is observed in the warehouse. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2169,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2170,tool_use,selection,medium,"Available tools: code_interpreter, calendar, terminal, database_query, browser.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2171,mathematical_reasoning,arithmetic,easy,Start with 28. Subtract 2. Then add 11. What is the final value? Show all steps.,"Step 1: 28 - 2 = 26
Step 2: 26 + 11 = 37
Final answer: 37","math,arithmetic,step-by-step"
arb_2172,logical_reasoning,deduction,medium,"If someone owns a wallet, then they have access to the garage. Avery owns a wallet. What can we conclude about Avery?","We can conclude that Avery has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_2173,logical_reasoning,deduction,easy,No Rileys are in the warehouse. Alex is in the warehouse. Is Alex a Riley? Explain.,"No. The premise states that no Rileys are in the warehouse. Since Alex is in the warehouse, Alex cannot be a Riley.","deduction,syllogism,logic"
arb_2174,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the book from the library and deliver it to the office. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the book (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the office.
6. Deliver / place the book in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2175,commonsense_reasoning,everyday,easy,A purple bird is observed in the kitchen. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2176,tool_use,selection,medium,"Available tools: web_search, email_client, file_system, browser, calendar.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2177,commonsense_reasoning,everyday,easy,A purple dog is observed in the garage. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2178,logical_reasoning,deduction,easy,"Either Taylor or Morgan took the laptop, but not both. Taylor did not take the laptop. Who took it?",Morgan took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2179,commonsense_reasoning,everyday,easy,A yellow rabbit is observed in the garage. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2180,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2181,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2182,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the phone from the kitchen and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the phone (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the office.
6. Deliver / place the phone in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2183,logical_reasoning,deduction,medium,"If someone owns a phone, then they have access to the meeting room. Jamie owns a phone. What can we conclude about Jamie?","We can conclude that Jamie has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_2184,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2185,tool_use,selection,medium,"Available tools: email_client, database_query, calendar, file_system, terminal.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2186,logical_reasoning,deduction,easy,All Rileys own a key. Taylor is a Riley. Does Taylor own a key? Explain step by step.,"Yes. From the universal statement 'All Rileys own a key' and the particular 'Taylor is a Riley', it follows by universal instantiation that Taylor owns a key.","deduction,syllogism,logic"
arb_2187,mathematical_reasoning,arithmetic,easy,Compute (45 * 44) + 35. Explain the order of operations.,"Multiplication first: 45 * 44 = 1980
Then addition: 1980 + 35 = 2015
Final answer: 2015","math,arithmetic,step-by-step"
arb_2188,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2189,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the bedroom. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2190,logical_reasoning,deduction,medium,"If someone owns a book, then they have access to the library. Cameron owns a book. What can we conclude about Cameron?","We can conclude that Cameron has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_2191,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the notebook from the server room and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the notebook (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the meeting room.
6. Deliver / place the notebook in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2192,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2193,mathematical_reasoning,arithmetic,easy,Start with 41. Subtract 26. Then add 8. What is the final value? Show all steps.,"Step 1: 41 - 26 = 15
Step 2: 15 + 8 = 23
Final answer: 23","math,arithmetic,step-by-step"
arb_2194,tool_use,selection,easy,"Available tools: code_interpreter, database_query, browser.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2195,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the book from the server room and deliver it to the library. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the book (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the library.
6. Deliver / place the book in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2196,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2197,mathematical_reasoning,arithmetic,easy,Compute (6 * 40) + 43. Explain the order of operations.,"Multiplication first: 6 * 40 = 240
Then addition: 240 + 43 = 283
Final answer: 283","math,arithmetic,step-by-step"
arb_2198,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2199,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2200,mathematical_reasoning,arithmetic,easy,Start with 33. Add 8. Then multiply the result by 47. What is the final value? Show all steps.,"Step 1: 33 + 8 = 41
Step 2: 41 * 47 = 1927
Final answer: 1927","math,arithmetic,step-by-step"
arb_2201,commonsense_reasoning,everyday,easy,A blue dog is observed in the kitchen. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2202,commonsense_reasoning,everyday,easy,A orange fox is observed in the kitchen. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2203,tool_use,selection,medium,"Available tools: terminal, database_query, file_system.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2204,agentic_planning,multi_step,medium,You are an agent assisting Quinn. Goal: retrieve the key from the lobby and deliver it to the meeting room. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2205,logical_reasoning,deduction,easy,No Averys are in the bedroom. Cameron is in the bedroom. Is Cameron a Avery? Explain.,"No. The premise states that no Averys are in the bedroom. Since Cameron is in the bedroom, Cameron cannot be a Avery.","deduction,syllogism,logic"
arb_2206,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2207,mathematical_reasoning,arithmetic,easy,Compute (1 * 7) + 41. Explain the order of operations.,"Multiplication first: 1 * 7 = 7
Then addition: 7 + 41 = 48
Final answer: 48","math,arithmetic,step-by-step"
arb_2208,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_2209,tool_use,selection,easy,"Available tools: browser, code_interpreter, calendar, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2210,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2211,tool_use,selection,medium,"Available tools: web_search, calendar, browser, email_client, code_interpreter.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2212,commonsense_reasoning,everyday,easy,A white rabbit is observed in the office. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a office (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2213,mathematical_reasoning,arithmetic,easy,Start with 8. Add 21. Then multiply the result by 3. What is the final value? Show all steps.,"Step 1: 8 + 21 = 29
Step 2: 29 * 3 = 87
Final answer: 87","math,arithmetic,step-by-step"
arb_2214,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2215,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the notebook from the garage and deliver it to the lobby. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the notebook (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the lobby.
6. Deliver / place the notebook in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2216,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2217,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_2218,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2219,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2220,mathematical_reasoning,arithmetic,easy,Start with 39. Subtract 25. Then add 6. What is the final value? Show all steps.,"Step 1: 39 - 25 = 14
Step 2: 14 + 6 = 20
Final answer: 20","math,arithmetic,step-by-step"
arb_2221,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the charger from the bedroom and deliver it to the garage. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the garage.
6. Deliver / place the charger in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2222,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2223,logical_reasoning,deduction,easy,All Morgans own a charger. Jamie is a Morgan. Does Jamie own a charger? Explain step by step.,"Yes. From the universal statement 'All Morgans own a charger' and the particular 'Jamie is a Morgan', it follows by universal instantiation that Jamie owns a charger.","deduction,syllogism,logic"
arb_2224,tool_use,selection,medium,"Available tools: calendar, file_system, terminal, browser, web_search.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2225,mathematical_reasoning,arithmetic,easy,Start with 39. Add 46. Then multiply the result by 48. What is the final value? Show all steps.,"Step 1: 39 + 46 = 85
Step 2: 85 * 48 = 4080
Final answer: 4080","math,arithmetic,step-by-step"
arb_2226,logical_reasoning,deduction,easy,No Jordans are in the library. Riley is in the library. Is Riley a Jordan? Explain.,"No. The premise states that no Jordans are in the library. Since Riley is in the library, Riley cannot be a Jordan.","deduction,syllogism,logic"
arb_2227,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2228,agentic_planning,multi_step,medium,You are an agent assisting Taylor. Goal: retrieve the pen from the office and deliver it to the lab. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the pen (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the lab.
6. Deliver / place the pen in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2229,tool_use,selection,medium,"Available tools: browser, file_system, terminal.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2230,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the pen from the warehouse and deliver it to the server room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the pen (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2231,tool_use,selection,medium,"Available tools: browser, calendar, database_query, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2232,logical_reasoning,deduction,medium,No Camerons are in the meeting room. Riley is in the meeting room. Is Riley a Cameron? Explain.,"No. The premise states that no Camerons are in the meeting room. Since Riley is in the meeting room, Riley cannot be a Cameron.","deduction,syllogism,logic"
arb_2233,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the pen from the kitchen and deliver it to the warehouse. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the pen (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the warehouse.
6. Deliver / place the pen in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2234,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2235,commonsense_reasoning,everyday,easy,A yellow owl is observed in the warehouse. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2236,commonsense_reasoning,everyday,easy,A blue bird is observed in the garage. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2237,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2238,mathematical_reasoning,arithmetic,easy,Start with 6. Add 2. Then multiply the result by 14. What is the final value? Show all steps.,"Step 1: 6 + 2 = 8
Step 2: 8 * 14 = 112
Final answer: 112","math,arithmetic,step-by-step"
arb_2239,tool_use,selection,medium,"Available tools: code_interpreter, browser, terminal, calendar, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2240,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2241,tool_use,selection,easy,"Available tools: web_search, calendar, database_query, email_client, browser.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2242,mathematical_reasoning,arithmetic,easy,Start with 23. Add 44. Then multiply the result by 11. What is the final value? Show all steps.,"Step 1: 23 + 44 = 67
Step 2: 67 * 11 = 737
Final answer: 737","math,arithmetic,step-by-step"
arb_2243,tool_use,selection,easy,"Available tools: file_system, code_interpreter, terminal, database_query.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2244,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2245,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_2246,logical_reasoning,deduction,medium,No Morgans are in the server room. Jamie is in the server room. Is Jamie a Morgan? Explain.,"No. The premise states that no Morgans are in the server room. Since Jamie is in the server room, Jamie cannot be a Morgan.","deduction,syllogism,logic"
arb_2247,agentic_planning,multi_step,hard,You are an agent assisting Riley. Goal: retrieve the charger from the bedroom and deliver it to the lab. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the charger (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the charger.
4. Verify possession of the charger (success check).
5. Navigate to the lab.
6. Deliver / place the charger in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2248,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_2249,mathematical_reasoning,arithmetic,easy,Start with 27. Subtract 5. Then add 16. What is the final value? Show all steps.,"Step 1: 27 - 5 = 22
Step 2: 22 + 16 = 38
Final answer: 38","math,arithmetic,step-by-step"
arb_2250,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2251,logical_reasoning,deduction,easy,"If someone owns a phone, then they have access to the kitchen. Jordan owns a phone. What can we conclude about Jordan?","We can conclude that Jordan has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_2252,mathematical_reasoning,arithmetic,easy,Compute (37 * 15) + 34. Explain the order of operations.,"Multiplication first: 37 * 15 = 555
Then addition: 555 + 34 = 589
Final answer: 589","math,arithmetic,step-by-step"
arb_2253,commonsense_reasoning,everyday,easy,A black cat is observed in the garage. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2254,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2255,commonsense_reasoning,everyday,easy,A green fox is observed in the office. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2256,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_2257,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the umbrella from the server room and deliver it to the warehouse. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the umbrella (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the warehouse.
6. Deliver / place the umbrella in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2258,tool_use,selection,medium,"Available tools: browser, web_search, file_system.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2259,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2260,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_2261,logical_reasoning,deduction,medium,"If someone owns a badge, then they have access to the warehouse. Taylor owns a badge. What can we conclude about Taylor?","We can conclude that Taylor has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_2262,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2263,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the umbrella from the warehouse and deliver it to the library. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the umbrella (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the library.
6. Deliver / place the umbrella in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2264,mathematical_reasoning,arithmetic,easy,Start with 39. Add 17. Then multiply the result by 5. What is the final value? Show all steps.,"Step 1: 39 + 17 = 56
Step 2: 56 * 5 = 280
Final answer: 280","math,arithmetic,step-by-step"
arb_2265,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the service downtime.","causal,diagnosis,agentic"
arb_2266,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2267,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2268,commonsense_reasoning,everyday,easy,A black bird is observed in the lobby. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2269,logical_reasoning,deduction,easy,"If someone owns a notebook, then they have access to the lab. Avery owns a notebook. What can we conclude about Avery?","We can conclude that Avery has access to the lab, by modus ponens.","deduction,syllogism,logic"
arb_2270,mathematical_reasoning,arithmetic,easy,Compute (25 * 20) + 26. Explain the order of operations.,"Multiplication first: 25 * 20 = 500
Then addition: 500 + 26 = 526
Final answer: 526","math,arithmetic,step-by-step"
arb_2271,tool_use,selection,medium,"Available tools: browser, email_client, file_system, database_query.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: browser
Approximate call: browser(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2272,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2273,logical_reasoning,deduction,easy,"Either Avery or Quinn took the notebook, but not both. Avery did not take the notebook. Who took it?",Quinn took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2274,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2275,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the key from the lobby and deliver it to the kitchen. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the kitchen.
6. Deliver / place the key in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2276,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the wallet from the kitchen and deliver it to the library. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the library.
6. Deliver / place the wallet in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2277,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2278,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the umbrella from the library and deliver it to the kitchen. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the umbrella (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the kitchen.
6. Deliver / place the umbrella in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2279,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the wallet from the warehouse and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the server room.
6. Deliver / place the wallet in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2280,mathematical_reasoning,arithmetic,easy,Compute (19 * 43) + 46. Explain the order of operations.,"Multiplication first: 19 * 43 = 817
Then addition: 817 + 46 = 863
Final answer: 863","math,arithmetic,step-by-step"
arb_2281,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the laptop from the kitchen and deliver it to the warehouse. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2282,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2283,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2284,logical_reasoning,deduction,medium,"Either Casey or Cameron took the pen, but not both. Casey did not take the pen. Who took it?",Cameron took the pen. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2285,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2286,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2287,tool_use,selection,easy,"Available tools: file_system, database_query, browser, calendar.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2288,commonsense_reasoning,everyday,easy,A orange dog is observed in the office. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2289,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2290,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2291,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_2292,tool_use,selection,easy,"Available tools: web_search, code_interpreter, terminal, browser, database_query.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2293,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the notebook from the lab and deliver it to the kitchen. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the notebook (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the kitchen.
6. Deliver / place the notebook in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2294,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the laptop from the kitchen and deliver it to the warehouse. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the warehouse.
6. Deliver / place the laptop in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2295,commonsense_reasoning,everyday,easy,A black rabbit is observed in the garage. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2296,agentic_planning,multi_step,hard,You are an agent assisting Alex. Goal: retrieve the phone from the server room and deliver it to the warehouse. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the phone (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the warehouse.
6. Deliver / place the phone in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2297,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2298,mathematical_reasoning,arithmetic,easy,Start with 24. Add 37. Then multiply the result by 19. What is the final value? Show all steps.,"Step 1: 24 + 37 = 61
Step 2: 61 * 19 = 1159
Final answer: 1159","math,arithmetic,step-by-step"
arb_2299,mathematical_reasoning,arithmetic,easy,Start with 8. Add 36. Then multiply the result by 23. What is the final value? Show all steps.,"Step 1: 8 + 36 = 44
Step 2: 44 * 23 = 1012
Final answer: 1012","math,arithmetic,step-by-step"
arb_2300,commonsense_reasoning,everyday,easy,A green rabbit is observed in the office. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a office (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2301,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Casey. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2302,tool_use,selection,medium,"Available tools: code_interpreter, database_query, terminal, calendar, file_system.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2303,mathematical_reasoning,arithmetic,easy,Start with 26. Subtract 7. Then add 40. What is the final value? Show all steps.,"Step 1: 26 - 7 = 19
Step 2: 19 + 40 = 59
Final answer: 59","math,arithmetic,step-by-step"
arb_2304,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the book from the server room and deliver it to the kitchen. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the book (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the kitchen.
6. Deliver / place the book in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2305,commonsense_reasoning,everyday,easy,A red rabbit is observed in the bedroom. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2306,tool_use,selection,medium,"Available tools: database_query, code_interpreter, calendar, terminal, email_client.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2307,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_2308,commonsense_reasoning,everyday,easy,A red fox is observed in the library. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2309,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2310,logical_reasoning,deduction,medium,All Morgans own a laptop. Casey is a Morgan. Does Casey own a laptop? Explain step by step.,"Yes. From the universal statement 'All Morgans own a laptop' and the particular 'Casey is a Morgan', it follows by universal instantiation that Casey owns a laptop.","deduction,syllogism,logic"
arb_2311,tool_use,selection,medium,"Available tools: calendar, file_system, browser, terminal, web_search.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2312,logical_reasoning,deduction,easy,No Alexs are in the server room. Jamie is in the server room. Is Jamie a Alex? Explain.,"No. The premise states that no Alexs are in the server room. Since Jamie is in the server room, Jamie cannot be a Alex.","deduction,syllogism,logic"
arb_2313,commonsense_reasoning,everyday,easy,A green bird is observed in the warehouse. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2314,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2315,logical_reasoning,deduction,easy,"If someone owns a phone, then they have access to the garage. Taylor owns a phone. What can we conclude about Taylor?","We can conclude that Taylor has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_2316,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2317,tool_use,selection,easy,"Available tools: code_interpreter, database_query, calendar, email_client, file_system.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2318,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2319,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2320,logical_reasoning,deduction,medium,"Either Jordan or Alex took the notebook, but not both. Jordan did not take the notebook. Who took it?",Alex took the notebook. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2321,mathematical_reasoning,arithmetic,easy,Start with 40. Add 7. Then multiply the result by 36. What is the final value? Show all steps.,"Step 1: 40 + 7 = 47
Step 2: 47 * 36 = 1692
Final answer: 1692","math,arithmetic,step-by-step"
arb_2322,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2323,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2324,tool_use,selection,easy,"Available tools: web_search, database_query, email_client, calendar.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2325,commonsense_reasoning,everyday,easy,A green dog is observed in the garage. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2326,mathematical_reasoning,arithmetic,easy,Start with 49. Add 33. Then multiply the result by 25. What is the final value? Show all steps.,"Step 1: 49 + 33 = 82
Step 2: 82 * 25 = 2050
Final answer: 2050","math,arithmetic,step-by-step"
arb_2327,mathematical_reasoning,arithmetic,easy,Start with 5. Add 3. Then multiply the result by 33. What is the final value? Show all steps.,"Step 1: 5 + 3 = 8
Step 2: 8 * 33 = 264
Final answer: 264","math,arithmetic,step-by-step"
arb_2328,commonsense_reasoning,everyday,easy,A red cat is observed in the office. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a office (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2329,commonsense_reasoning,everyday,easy,A blue bird is observed in the kitchen. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2330,tool_use,selection,easy,"Available tools: code_interpreter, web_search, browser, email_client, calendar.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2331,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2332,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2333,agentic_planning,multi_step,medium,You are an agent assisting Jamie. Goal: retrieve the wallet from the library and deliver it to the office. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the office.
6. Deliver / place the wallet in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2334,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2335,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2336,agentic_planning,multi_step,medium,You are an agent assisting Casey. Goal: retrieve the phone from the library and deliver it to the server room. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the phone (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the server room.
6. Deliver / place the phone in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2337,mathematical_reasoning,arithmetic,easy,Start with 21. Add 25. Then multiply the result by 17. What is the final value? Show all steps.,"Step 1: 21 + 25 = 46
Step 2: 46 * 17 = 782
Final answer: 782","math,arithmetic,step-by-step"
arb_2338,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the key from the office and deliver it to the lobby. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the key (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lobby.
6. Deliver / place the key in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2339,agentic_planning,multi_step,medium,You are an agent assisting Jordan. Goal: retrieve the key from the warehouse and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the key (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the garage.
6. Deliver / place the key in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2340,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2341,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the laptop from the library and deliver it to the meeting room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the laptop (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the meeting room.
6. Deliver / place the laptop in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2342,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2343,mathematical_reasoning,arithmetic,easy,Start with 13. Add 10. Then multiply the result by 4. What is the final value? Show all steps.,"Step 1: 13 + 10 = 23
Step 2: 23 * 4 = 92
Final answer: 92","math,arithmetic,step-by-step"
arb_2344,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2345,mathematical_reasoning,arithmetic,easy,Start with 42. Add 5. Then multiply the result by 35. What is the final value? Show all steps.,"Step 1: 42 + 5 = 47
Step 2: 47 * 35 = 1645
Final answer: 1645","math,arithmetic,step-by-step"
arb_2346,mathematical_reasoning,arithmetic,easy,Start with 27. Subtract 10. Then add 2. What is the final value? Show all steps.,"Step 1: 27 - 10 = 17
Step 2: 17 + 2 = 19
Final answer: 19","math,arithmetic,step-by-step"
arb_2347,logical_reasoning,deduction,medium,No Averys are in the library. Casey is in the library. Is Casey a Avery? Explain.,"No. The premise states that no Averys are in the library. Since Casey is in the library, Casey cannot be a Avery.","deduction,syllogism,logic"
arb_2348,mathematical_reasoning,arithmetic,easy,Start with 36. Subtract 20. Then add 34. What is the final value? Show all steps.,"Step 1: 36 - 20 = 16
Step 2: 16 + 34 = 50
Final answer: 50","math,arithmetic,step-by-step"
arb_2349,logical_reasoning,deduction,medium,"If someone owns a charger, then they have access to the office. Alex owns a charger. What can we conclude about Alex?","We can conclude that Alex has access to the office, by modus ponens.","deduction,syllogism,logic"
arb_2350,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the key from the bedroom and deliver it to the lobby. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the key (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the lobby.
6. Deliver / place the key in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2351,logical_reasoning,deduction,easy,All Jamies own a badge. Cameron is a Jamie. Does Cameron own a badge? Explain step by step.,"Yes. From the universal statement 'All Jamies own a badge' and the particular 'Cameron is a Jamie', it follows by universal instantiation that Cameron owns a badge.","deduction,syllogism,logic"
arb_2352,mathematical_reasoning,arithmetic,easy,Start with 48. Add 9. Then multiply the result by 17. What is the final value? Show all steps.,"Step 1: 48 + 9 = 57
Step 2: 57 * 17 = 969
Final answer: 969","math,arithmetic,step-by-step"
arb_2353,logical_reasoning,deduction,medium,"Either Jamie or Taylor took the laptop, but not both. Jamie did not take the laptop. Who took it?",Taylor took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2354,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2355,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Quinn. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2356,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2357,tool_use,selection,medium,"Available tools: code_interpreter, terminal, email_client.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2358,agentic_planning,multi_step,hard,You are an agent assisting Cameron. Goal: retrieve the umbrella from the garage and deliver it to the lobby. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the umbrella (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the lobby.
6. Deliver / place the umbrella in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2359,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the laptop from the kitchen and deliver it to the lab. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the laptop (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2360,tool_use,selection,medium,"Available tools: web_search, terminal, browser.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: email_client
Approximate call: email_client(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2361,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2362,mathematical_reasoning,arithmetic,easy,Compute (2 * 18) + 36. Explain the order of operations.,"Multiplication first: 2 * 18 = 36
Then addition: 36 + 36 = 72
Final answer: 72","math,arithmetic,step-by-step"
arb_2363,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the notebook from the lab and deliver it to the garage. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the notebook (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the garage.
6. Deliver / place the notebook in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2364,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the pen from the lab and deliver it to the server room. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the pen (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the server room.
6. Deliver / place the pen in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2365,logical_reasoning,deduction,easy,All Jamies own a key. Morgan is a Jamie. Does Morgan own a key? Explain step by step.,"Yes. From the universal statement 'All Jamies own a key' and the particular 'Morgan is a Jamie', it follows by universal instantiation that Morgan owns a key.","deduction,syllogism,logic"
arb_2366,logical_reasoning,deduction,medium,"Either Riley or Taylor took the laptop, but not both. Riley did not take the laptop. Who took it?",Taylor took the laptop. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2367,tool_use,selection,easy,"Available tools: file_system, code_interpreter, calendar, email_client.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2368,logical_reasoning,deduction,medium,All Averys own a wallet. Jordan is a Avery. Does Jordan own a wallet? Explain step by step.,"Yes. From the universal statement 'All Averys own a wallet' and the particular 'Jordan is a Avery', it follows by universal instantiation that Jordan owns a wallet.","deduction,syllogism,logic"
arb_2369,agentic_planning,multi_step,hard,You are an agent assisting Casey. Goal: retrieve the phone from the office and deliver it to the server room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the phone (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the server room.
6. Deliver / place the phone in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2370,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2371,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the pen from the garage and deliver it to the kitchen. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the pen (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the kitchen.
6. Deliver / place the pen in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2372,commonsense_reasoning,everyday,easy,A blue owl is observed in the lobby. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2373,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2374,mathematical_reasoning,arithmetic,easy,Start with 11. Subtract 6. Then add 18. What is the final value? Show all steps.,"Step 1: 11 - 6 = 5
Step 2: 5 + 18 = 23
Final answer: 23","math,arithmetic,step-by-step"
arb_2375,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_2376,logical_reasoning,deduction,easy,No Jamies are in the lab. Morgan is in the lab. Is Morgan a Jamie? Explain.,"No. The premise states that no Jamies are in the lab. Since Morgan is in the lab, Morgan cannot be a Jamie.","deduction,syllogism,logic"
arb_2377,commonsense_reasoning,everyday,easy,A black dog is observed in the lab. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2378,mathematical_reasoning,arithmetic,easy,Start with 12. Add 48. Then multiply the result by 32. What is the final value? Show all steps.,"Step 1: 12 + 48 = 60
Step 2: 60 * 32 = 1920
Final answer: 1920","math,arithmetic,step-by-step"
arb_2379,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Morgan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2380,logical_reasoning,deduction,easy,"Either Casey or Cameron took the charger, but not both. Casey did not take the charger. Who took it?",Cameron took the charger. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2381,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the user complaints.","causal,diagnosis,agentic"
arb_2382,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_2383,mathematical_reasoning,arithmetic,easy,Start with 37. Subtract 34. Then add 13. What is the final value? Show all steps.,"Step 1: 37 - 34 = 3
Step 2: 3 + 13 = 16
Final answer: 16","math,arithmetic,step-by-step"
arb_2384,mathematical_reasoning,arithmetic,easy,Start with 47. Add 37. Then multiply the result by 9. What is the final value? Show all steps.,"Step 1: 47 + 37 = 84
Step 2: 84 * 9 = 756
Final answer: 756","math,arithmetic,step-by-step"
arb_2385,mathematical_reasoning,arithmetic,easy,Compute (43 * 42) + 10. Explain the order of operations.,"Multiplication first: 43 * 42 = 1806
Then addition: 1806 + 10 = 1816
Final answer: 1816","math,arithmetic,step-by-step"
arb_2386,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2387,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the book from the lab and deliver it to the server room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the book (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the book.
4. Verify possession of the book (success check).
5. Navigate to the server room.
6. Deliver / place the book in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2388,commonsense_reasoning,everyday,easy,A yellow cat is observed in the kitchen. Is it more likely that the cat lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about cats.,"Most cats do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the cat is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2389,logical_reasoning,deduction,medium,"If someone owns a phone, then they have access to the garage. Casey owns a phone. What can we conclude about Casey?","We can conclude that Casey has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_2390,logical_reasoning,deduction,easy,"Either Morgan or Taylor took the book, but not both. Morgan did not take the book. Who took it?",Taylor took the book. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2391,mathematical_reasoning,arithmetic,easy,Start with 5. Add 41. Then multiply the result by 7. What is the final value? Show all steps.,"Step 1: 5 + 41 = 46
Step 2: 46 * 7 = 322
Final answer: 322","math,arithmetic,step-by-step"
arb_2392,logical_reasoning,deduction,easy,"If someone owns a key, then they have access to the library. Alex owns a key. What can we conclude about Alex?","We can conclude that Alex has access to the library, by modus ponens.","deduction,syllogism,logic"
arb_2393,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2394,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2395,logical_reasoning,deduction,easy,"If someone owns a badge, then they have access to the server room. Avery owns a badge. What can we conclude about Avery?","We can conclude that Avery has access to the server room, by modus ponens.","deduction,syllogism,logic"
arb_2396,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_2397,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the wallet from the server room and deliver it to the office. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the office.
6. Deliver / place the wallet in the office and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2398,mathematical_reasoning,arithmetic,easy,Compute (26 * 43) + 48. Explain the order of operations.,"Multiplication first: 26 * 43 = 1118
Then addition: 1118 + 48 = 1166
Final answer: 1166","math,arithmetic,step-by-step"
arb_2399,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2400,logical_reasoning,deduction,easy,"Either Riley or Taylor took the badge, but not both. Riley did not take the badge. Who took it?",Taylor took the badge. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2401,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2402,logical_reasoning,deduction,medium,No Rileys are in the library. Jordan is in the library. Is Jordan a Riley? Explain.,"No. The premise states that no Rileys are in the library. Since Jordan is in the library, Jordan cannot be a Riley.","deduction,syllogism,logic"
arb_2403,mathematical_reasoning,arithmetic,easy,Start with 45. Subtract 44. Then add 26. What is the final value? Show all steps.,"Step 1: 45 - 44 = 1
Step 2: 1 + 26 = 27
Final answer: 27","math,arithmetic,step-by-step"
arb_2404,logical_reasoning,deduction,medium,"Either Casey or Jordan took the umbrella, but not both. Casey did not take the umbrella. Who took it?",Jordan took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2405,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2406,tool_use,selection,easy,"Available tools: calendar, email_client, database_query, browser.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2407,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2408,mathematical_reasoning,arithmetic,easy,Start with 14. Subtract 40. Then add 33. What is the final value? Show all steps.,"Step 1: 14 - 40 = -26
Step 2: -26 + 33 = 7
Final answer: 7","math,arithmetic,step-by-step"
arb_2409,tool_use,selection,easy,"Available tools: database_query, terminal, web_search.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2410,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2411,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the badge from the library and deliver it to the meeting room. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the badge (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the meeting room.
6. Deliver / place the badge in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2412,agentic_planning,multi_step,hard,You are an agent assisting Jamie. Goal: retrieve the pen from the lobby and deliver it to the garage. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the pen (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the garage.
6. Deliver / place the pen in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2413,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2414,mathematical_reasoning,arithmetic,easy,Compute (40 * 29) + 21. Explain the order of operations.,"Multiplication first: 40 * 29 = 1160
Then addition: 1160 + 21 = 1181
Final answer: 1181","math,arithmetic,step-by-step"
arb_2415,logical_reasoning,deduction,easy,No Quinns are in the lab. Riley is in the lab. Is Riley a Quinn? Explain.,"No. The premise states that no Quinns are in the lab. Since Riley is in the lab, Riley cannot be a Quinn.","deduction,syllogism,logic"
arb_2416,commonsense_reasoning,everyday,easy,A purple fish is observed in the office. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2417,commonsense_reasoning,everyday,easy,A white dog is observed in the office. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a office (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2418,logical_reasoning,deduction,medium,"If someone owns a charger, then they have access to the warehouse. Riley owns a charger. What can we conclude about Riley?","We can conclude that Riley has access to the warehouse, by modus ponens.","deduction,syllogism,logic"
arb_2419,tool_use,selection,easy,"Available tools: web_search, terminal, code_interpreter, email_client.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2420,tool_use,selection,medium,"Available tools: code_interpreter, browser, file_system.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2421,mathematical_reasoning,arithmetic,easy,Compute (44 * 11) + 4. Explain the order of operations.,"Multiplication first: 44 * 11 = 484
Then addition: 484 + 4 = 488
Final answer: 488","math,arithmetic,step-by-step"
arb_2422,mathematical_reasoning,arithmetic,easy,Start with 35. Subtract 12. Then add 2. What is the final value? Show all steps.,"Step 1: 35 - 12 = 23
Step 2: 23 + 2 = 25
Final answer: 25","math,arithmetic,step-by-step"
arb_2423,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2424,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
for i in range(5):
print(i)
```
Explain the error and provide a corrected version.","IndentationError: the print statement must be indented under the for-loop.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2425,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2426,logical_reasoning,deduction,easy,All Alexs own a laptop. Quinn is a Alex. Does Quinn own a laptop? Explain step by step.,"Yes. From the universal statement 'All Alexs own a laptop' and the particular 'Quinn is a Alex', it follows by universal instantiation that Quinn owns a laptop.","deduction,syllogism,logic"
arb_2427,mathematical_reasoning,arithmetic,easy,Start with 39. Subtract 28. Then add 6. What is the final value? Show all steps.,"Step 1: 39 - 28 = 11
Step 2: 11 + 6 = 17
Final answer: 17","math,arithmetic,step-by-step"
arb_2428,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Taylor. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2429,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2430,tool_use,selection,easy,"Available tools: database_query, terminal, code_interpreter.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2431,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2432,tool_use,selection,medium,"Available tools: terminal, calendar, database_query, web_search, browser.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2433,mathematical_reasoning,arithmetic,easy,Start with 28. Add 35. Then multiply the result by 2. What is the final value? Show all steps.,"Step 1: 28 + 35 = 63
Step 2: 63 * 2 = 126
Final answer: 126","math,arithmetic,step-by-step"
arb_2434,causal_reasoning,diagnosis,medium,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2435,logical_reasoning,deduction,medium,"If someone owns a notebook, then they have access to the meeting room. Cameron owns a notebook. What can we conclude about Cameron?","We can conclude that Cameron has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_2436,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jamie. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2437,tool_use,selection,medium,"Available tools: file_system, calendar, database_query, web_search, browser.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2438,mathematical_reasoning,arithmetic,easy,Compute (38 * 16) + 32. Explain the order of operations.,"Multiplication first: 38 * 16 = 608
Then addition: 608 + 32 = 640
Final answer: 640","math,arithmetic,step-by-step"
arb_2439,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2440,commonsense_reasoning,everyday,easy,A purple bird is observed in the office. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a office (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2441,tool_use,selection,easy,"Available tools: email_client, calendar, terminal, web_search, browser.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2442,mathematical_reasoning,arithmetic,easy,Start with 49. Add 26. Then multiply the result by 35. What is the final value? Show all steps.,"Step 1: 49 + 26 = 75
Step 2: 75 * 35 = 2625
Final answer: 2625","math,arithmetic,step-by-step"
arb_2443,logical_reasoning,deduction,easy,All Jamies own a key. Casey is a Jamie. Does Casey own a key? Explain step by step.,"Yes. From the universal statement 'All Jamies own a key' and the particular 'Casey is a Jamie', it follows by universal instantiation that Casey owns a key.","deduction,syllogism,logic"
arb_2444,mathematical_reasoning,arithmetic,easy,Start with 21. Add 26. Then multiply the result by 38. What is the final value? Show all steps.,"Step 1: 21 + 26 = 47
Step 2: 47 * 38 = 1786
Final answer: 1786","math,arithmetic,step-by-step"
arb_2445,logical_reasoning,deduction,medium,"If someone owns a badge, then they have access to the meeting room. Jordan owns a badge. What can we conclude about Jordan?","We can conclude that Jordan has access to the meeting room, by modus ponens.","deduction,syllogism,logic"
arb_2446,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the notebook from the lobby and deliver it to the garage. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the notebook (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the garage.
6. Deliver / place the notebook in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2447,mathematical_reasoning,arithmetic,easy,Compute (9 * 49) + 33. Explain the order of operations.,"Multiplication first: 9 * 49 = 441
Then addition: 441 + 33 = 474
Final answer: 474","math,arithmetic,step-by-step"
arb_2448,agentic_planning,multi_step,hard,You are an agent assisting Taylor. Goal: retrieve the wallet from the warehouse and deliver it to the server room. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the wallet (confirm it is in warehouse).
2. Navigate or request access to the warehouse.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the server room.
6. Deliver / place the wallet in the server room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2449,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2450,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the error logs.","causal,diagnosis,agentic"
arb_2451,tool_use,selection,easy,"Available tools: database_query, web_search, code_interpreter, calendar, email_client.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2452,logical_reasoning,deduction,easy,"If someone owns a laptop, then they have access to the lobby. Morgan owns a laptop. What can we conclude about Morgan?","We can conclude that Morgan has access to the lobby, by modus ponens.","deduction,syllogism,logic"
arb_2453,tool_use,selection,easy,"Available tools: database_query, code_interpreter, terminal, browser.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2454,causal_reasoning,diagnosis,hard,Observed effect: user complaints. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the user complaints.","causal,diagnosis,agentic"
arb_2455,tool_use,selection,medium,"Available tools: file_system, web_search, terminal, email_client, code_interpreter.
User goal: restart the nginx service.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(appropriate_arguments)
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2456,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2457,mathematical_reasoning,arithmetic,easy,Start with 2. Add 27. Then multiply the result by 41. What is the final value? Show all steps.,"Step 1: 2 + 27 = 29
Step 2: 29 * 41 = 1189
Final answer: 1189","math,arithmetic,step-by-step"
arb_2458,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2459,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the badge from the kitchen and deliver it to the library. Available tools include terminal and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use terminal or file_system to locate the current position of the badge (confirm it is in kitchen).
2. Navigate or request access to the kitchen.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the library.
6. Deliver / place the badge in the library and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2460,logical_reasoning,deduction,easy,No Caseys are in the kitchen. Jamie is in the kitchen. Is Jamie a Casey? Explain.,"No. The premise states that no Caseys are in the kitchen. Since Jamie is in the kitchen, Jamie cannot be a Casey.","deduction,syllogism,logic"
arb_2461,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_2462,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the data loss.","causal,diagnosis,agentic"
arb_2463,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Cameron. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2464,mathematical_reasoning,arithmetic,easy,Compute (20 * 16) + 14. Explain the order of operations.,"Multiplication first: 20 * 16 = 320
Then addition: 320 + 14 = 334
Final answer: 334","math,arithmetic,step-by-step"
arb_2465,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2466,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2467,tool_use,selection,easy,"Available tools: file_system, terminal, web_search, email_client.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2468,commonsense_reasoning,everyday,easy,A yellow dog is observed in the garage. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a garage (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2469,logical_reasoning,deduction,medium,No Taylors are in the warehouse. Jordan is in the warehouse. Is Jordan a Taylor? Explain.,"No. The premise states that no Taylors are in the warehouse. Since Jordan is in the warehouse, Jordan cannot be a Taylor.","deduction,syllogism,logic"
arb_2470,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the laptop from the server room and deliver it to the garage. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the laptop (confirm it is in server room).
2. Navigate or request access to the server room.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the garage.
6. Deliver / place the laptop in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2471,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
def add(a, b):
    return a - b
```
Explain the error and provide a corrected version.","The function is named 'add' but performs subtraction. It should return a + b.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2472,logical_reasoning,deduction,medium,"If someone owns a book, then they have access to the kitchen. Cameron owns a book. What can we conclude about Cameron?","We can conclude that Cameron has access to the kitchen, by modus ponens.","deduction,syllogism,logic"
arb_2473,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the data loss.","causal,diagnosis,agentic"
arb_2474,causal_reasoning,diagnosis,medium,Observed effect: service downtime. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the service downtime.","causal,diagnosis,agentic"
arb_2475,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2476,tool_use,selection,easy,"Available tools: web_search, database_query, calendar, file_system.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2477,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2478,causal_reasoning,diagnosis,medium,Observed effect: data loss. Possible causes include network failure and others. What evidence would support or refute that network failure is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'network failure': temporal correlation (cause preceded effect), presence of characteristic symptoms of network failure, absence of alternative explanations.
Refuting evidence: the effect occurred without network failure, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of network failure indicators.
2. Inspect logs for signatures of network failure.
3. Test whether removing/fixing network failure eliminates the data loss.","causal,diagnosis,agentic"
arb_2479,agentic_planning,multi_step,hard,You are an agent assisting Quinn. Goal: retrieve the notebook from the office and deliver it to the kitchen. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the notebook (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the kitchen.
6. Deliver / place the notebook in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2480,mathematical_reasoning,arithmetic,easy,Start with 32. Subtract 2. Then add 22. What is the final value? Show all steps.,"Step 1: 32 - 2 = 30
Step 2: 30 + 22 = 52
Final answer: 52","math,arithmetic,step-by-step"
arb_2481,mathematical_reasoning,arithmetic,easy,Start with 14. Add 19. Then multiply the result by 12. What is the final value? Show all steps.,"Step 1: 14 + 19 = 33
Step 2: 33 * 12 = 396
Final answer: 396","math,arithmetic,step-by-step"
arb_2482,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2483,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
x = [1, 2, 3]
print(x[3])
```
Explain the error and provide a corrected version.","IndexError: list index out of range. Valid indices are 0, 1, 2.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2484,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2485,code_reasoning,debugging,easy,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2486,commonsense_reasoning,everyday,easy,A green bird is observed in the warehouse. Is it more likely that the bird lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about birds.,"Most birds do not permanently live in a warehouse (which is a human-built environment). It is therefore more likely that the bird is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2487,tool_use,selection,medium,"Available tools: web_search, database_query, terminal, calendar.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2488,agentic_planning,multi_step,medium,You are an agent assisting Alex. Goal: retrieve the wallet from the library and deliver it to the garage. Available tools include code_interpreter and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use code_interpreter or file_system to locate the current position of the wallet (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the garage.
6. Deliver / place the wallet in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2489,mathematical_reasoning,arithmetic,easy,Start with 39. Add 27. Then multiply the result by 8. What is the final value? Show all steps.,"Step 1: 39 + 27 = 66
Step 2: 66 * 8 = 528
Final answer: 528","math,arithmetic,step-by-step"
arb_2490,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2491,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2492,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the notebook from the office and deliver it to the bedroom. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the notebook (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the bedroom.
6. Deliver / place the notebook in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2493,causal_reasoning,diagnosis,hard,Observed effect: data loss. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the data loss.","causal,diagnosis,agentic"
arb_2494,mathematical_reasoning,arithmetic,easy,Compute (30 * 43) + 47. Explain the order of operations.,"Multiplication first: 30 * 43 = 1290
Then addition: 1290 + 47 = 1337
Final answer: 1337","math,arithmetic,step-by-step"
arb_2495,tool_use,selection,easy,"Available tools: web_search, file_system, email_client, calendar.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2496,tool_use,selection,easy,"Available tools: email_client, web_search, terminal, database_query, file_system.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2497,tool_use,selection,easy,"Available tools: calendar, database_query, email_client.
User goal: query the sales table for Q3 revenue.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: database_query
Approximate call: database_query(sql=""SELECT SUM(revenue) FROM sales WHERE quarter='Q3'"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2498,tool_use,selection,easy,"Available tools: database_query, web_search, calendar, terminal.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2499,causal_reasoning,diagnosis,hard,Observed effect: error logs. Possible causes include power outage and others. What evidence would support or refute that power outage is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'power outage': temporal correlation (cause preceded effect), presence of characteristic symptoms of power outage, absence of alternative explanations.
Refuting evidence: the effect occurred without power outage, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of power outage indicators.
2. Inspect logs for signatures of power outage.
3. Test whether removing/fixing power outage eliminates the error logs.","causal,diagnosis,agentic"
arb_2500,commonsense_reasoning,everyday,easy,A yellow fish is observed in the meeting room. Is it more likely that the fish lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about fishs.,"Most fishs do not permanently live in a meeting room (which is a human-built environment). It is therefore more likely that the fish is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2501,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
d = {'a': 1}
print(d['b'])
```
Explain the error and provide a corrected version.","KeyError: 'b' is not a key in the dictionary. Use d.get('b') or check membership.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"
arb_2502,commonsense_reasoning,everyday,easy,A blue rabbit is observed in the kitchen. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2503,commonsense_reasoning,everyday,easy,A purple fox is observed in the kitchen. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a kitchen (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2504,mathematical_reasoning,arithmetic,easy,Start with 33. Subtract 3. Then add 5. What is the final value? Show all steps.,"Step 1: 33 - 3 = 30
Step 2: 30 + 5 = 35
Final answer: 35","math,arithmetic,step-by-step"
arb_2505,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2506,causal_reasoning,diagnosis,medium,Observed effect: user complaints. Possible causes include rain and others. What evidence would support or refute that rain is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'rain': temporal correlation (cause preceded effect), presence of characteristic symptoms of rain, absence of alternative explanations.
Refuting evidence: the effect occurred without rain, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of rain indicators.
2. Inspect logs for signatures of rain.
3. Test whether removing/fixing rain eliminates the user complaints.","causal,diagnosis,agentic"
arb_2507,agentic_planning,multi_step,medium,You are an agent assisting Avery. Goal: retrieve the notebook from the meeting room and deliver it to the warehouse. Available tools include web_search and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use web_search or file_system to locate the current position of the notebook (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the warehouse.
6. Deliver / place the notebook in the warehouse and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2508,tool_use,selection,easy,"Available tools: file_system, code_interpreter, terminal.
User goal: list all Python files in the current directory.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""ls *.py"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2509,commonsense_reasoning,everyday,easy,A yellow rabbit is observed in the library. Is it more likely that the rabbit lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about rabbits.,"Most rabbits do not permanently live in a library (which is a human-built environment). It is therefore more likely that the rabbit is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2510,mathematical_reasoning,arithmetic,easy,Start with 10. Subtract 2. Then add 6. What is the final value? Show all steps.,"Step 1: 10 - 2 = 8
Step 2: 8 + 6 = 14
Final answer: 14","math,arithmetic,step-by-step"
arb_2511,mathematical_reasoning,arithmetic,easy,Compute (9 * 6) + 21. Explain the order of operations.,"Multiplication first: 9 * 6 = 54
Then addition: 54 + 21 = 75
Final answer: 75","math,arithmetic,step-by-step"
arb_2512,agentic_planning,multi_step,hard,You are an agent assisting Jordan. Goal: retrieve the phone from the garage and deliver it to the lab. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the phone (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the phone.
4. Verify possession of the phone (success check).
5. Navigate to the lab.
6. Deliver / place the phone in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2513,logical_reasoning,deduction,medium,"Either Cameron or Jordan took the phone, but not both. Cameron did not take the phone. Who took it?",Jordan took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2514,tool_use,selection,easy,"Available tools: file_system, database_query, code_interpreter.
User goal: search for recent papers on transformer architectures.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: file_system
Approximate call: file_system(query=""search for recent papers on transformer architectures"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2515,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the umbrella from the garage and deliver it to the bedroom. Available tools include file_system and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use file_system or file_system to locate the current position of the umbrella (confirm it is in garage).
2. Navigate or request access to the garage.
3. Pick up / retrieve the umbrella.
4. Verify possession of the umbrella (success check).
5. Navigate to the bedroom.
6. Deliver / place the umbrella in the bedroom and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2516,mathematical_reasoning,arithmetic,easy,Start with 16. Subtract 43. Then add 21. What is the final value? Show all steps.,"Step 1: 16 - 43 = -27
Step 2: -27 + 21 = -6
Final answer: -6","math,arithmetic,step-by-step"
arb_2517,logical_reasoning,deduction,medium,No Jamies are in the kitchen. Avery is in the kitchen. Is Avery a Jamie? Explain.,"No. The premise states that no Jamies are in the kitchen. Since Avery is in the kitchen, Avery cannot be a Jamie.","deduction,syllogism,logic"
arb_2518,tool_use,selection,easy,"Available tools: calendar, terminal, code_interpreter, browser.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: code_interpreter
Approximate call: code_interpreter(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2519,agentic_planning,multi_step,hard,You are an agent assisting Avery. Goal: retrieve the laptop from the library and deliver it to the lab. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 5 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the laptop (confirm it is in library).
2. Navigate or request access to the library.
3. Pick up / retrieve the laptop.
4. Verify possession of the laptop (success check).
5. Navigate to the lab.
6. Deliver / place the laptop in the lab and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2520,causal_reasoning,diagnosis,hard,Observed effect: delayed delivery. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the delayed delivery.","causal,diagnosis,agentic"
arb_2521,tool_use,selection,easy,"Available tools: browser, email_client, terminal, web_search.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2522,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the key from the lobby and deliver it to the meeting room. Available tools include email_client and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use email_client or file_system to locate the current position of the key (confirm it is in lobby).
2. Navigate or request access to the lobby.
3. Pick up / retrieve the key.
4. Verify possession of the key (success check).
5. Navigate to the meeting room.
6. Deliver / place the key in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2523,commonsense_reasoning,everyday,easy,A orange fox is observed in the library. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a library (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2524,commonsense_reasoning,everyday,easy,A orange fox is observed in the lobby. Is it more likely that the fox lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about foxs.,"Most foxs do not permanently live in a lobby (which is a human-built environment). It is therefore more likely that the fox is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2525,logical_reasoning,deduction,easy,"Either Jamie or Morgan took the phone, but not both. Jamie did not take the phone. Who took it?",Morgan took the phone. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2526,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Riley. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2527,logical_reasoning,deduction,easy,"If someone owns a book, then they have access to the garage. Avery owns a book. What can we conclude about Avery?","We can conclude that Avery has access to the garage, by modus ponens.","deduction,syllogism,logic"
arb_2528,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2529,tool_use,selection,medium,"Available tools: email_client, calendar, terminal.
User goal: check disk usage on /var.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(command=""df -h /var"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2530,agentic_planning,multi_step,medium,You are an agent assisting Riley. Goal: retrieve the wallet from the bedroom and deliver it to the meeting room. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the wallet (confirm it is in bedroom).
2. Navigate or request access to the bedroom.
3. Pick up / retrieve the wallet.
4. Verify possession of the wallet (success check).
5. Navigate to the meeting room.
6. Deliver / place the wallet in the meeting room and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2531,tool_use,selection,easy,"Available tools: database_query, email_client, file_system, calendar.
User goal: send a meeting invitation for tomorrow at 10am.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: calendar
Approximate call: calendar(title=""Meeting"", time=""tomorrow 10:00"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2532,commonsense_reasoning,everyday,easy,A blue owl is observed in the bedroom. Is it more likely that the owl lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about owls.,"Most owls do not permanently live in a bedroom (which is a human-built environment). It is therefore more likely that the owl is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2533,agentic_planning,multi_step,hard,You are an agent assisting Morgan. Goal: retrieve the pen from the meeting room and deliver it to the garage. Available tools include browser and file_system. Provide a numbered step-by-step plan with at least 3 steps. Include checks for success and recovery actions.,"1. Use browser or file_system to locate the current position of the pen (confirm it is in meeting room).
2. Navigate or request access to the meeting room.
3. Pick up / retrieve the pen.
4. Verify possession of the pen (success check).
5. Navigate to the garage.
6. Deliver / place the pen in the garage and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2534,logical_reasoning,deduction,easy,No Jordans are in the kitchen. Casey is in the kitchen. Is Casey a Jordan? Explain.,"No. The premise states that no Jordans are in the kitchen. Since Casey is in the kitchen, Casey cannot be a Jordan.","deduction,syllogism,logic"
arb_2535,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Alex. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2536,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2537,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2538,commonsense_reasoning,everyday,easy,A red dog is observed in the lab. Is it more likely that the dog lives there permanently or is only visiting? Give a reasoned answer based on typical knowledge about dogs.,"Most dogs do not permanently live in a lab (which is a human-built environment). It is therefore more likely that the dog is only visiting or was brought there temporarily. Exceptions exist (e.g. pets), but the default assumption is temporary presence.","commonsense,plausibility"
arb_2539,causal_reasoning,diagnosis,hard,Observed effect: service downtime. Possible causes include software bug and others. What evidence would support or refute that software bug is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'software bug': temporal correlation (cause preceded effect), presence of characteristic symptoms of software bug, absence of alternative explanations.
Refuting evidence: the effect occurred without software bug, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of software bug indicators.
2. Inspect logs for signatures of software bug.
3. Test whether removing/fixing software bug eliminates the service downtime.","causal,diagnosis,agentic"
arb_2540,causal_reasoning,diagnosis,medium,Observed effect: error logs. Possible causes include user error and others. What evidence would support or refute that user error is the actual cause? Propose a simple diagnostic plan.,"Supporting evidence for 'user error': temporal correlation (cause preceded effect), presence of characteristic symptoms of user error, absence of alternative explanations.
Refuting evidence: the effect occurred without user error, or the effect is fully explained by another cause.
Diagnostic plan:
1. Check timestamps of user error indicators.
2. Inspect logs for signatures of user error.
3. Test whether removing/fixing user error eliminates the error logs.","causal,diagnosis,agentic"
arb_2541,agentic_planning,multi_step,medium,You are an agent assisting Morgan. Goal: retrieve the notebook from the office and deliver it to the kitchen. Available tools include database_query and file_system. Provide a numbered step-by-step plan with at least 4 steps. Include checks for success and recovery actions.,"1. Use database_query or file_system to locate the current position of the notebook (confirm it is in office).
2. Navigate or request access to the office.
3. Pick up / retrieve the notebook.
4. Verify possession of the notebook (success check).
5. Navigate to the kitchen.
6. Deliver / place the notebook in the kitchen and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2542,logical_reasoning,deduction,medium,"Either Avery or Jamie took the umbrella, but not both. Avery did not take the umbrella. Who took it?",Jamie took the umbrella. This follows from the exclusive disjunction and the negation of one disjunct.,"deduction,syllogism,logic"
arb_2543,agentic_interaction,clarification,medium,"User request: 'Get me the latest numbers.' The context is ambiguous (could mean financial metrics, sports scores, version numbers, etc.). How should an agent respond? Provide a clarification strategy and example clarifying questions.","An agent should not guess when critical ambiguity exists. Strategy: ask targeted clarifying questions while offering the most likely interpretations.
Example questions:
- Are you referring to financial / sales numbers, system metrics, or something else?
- For which time period and which entity?
- In which format do you need them (table, chart, raw values)?
After clarification, proceed with the appropriate tool calls.","ambiguity,clarification,user-interaction"
arb_2544,agentic_planning,long_horizon,hard,"You are a long-horizon agent. Your ultimate goal is to prepare a weekly status report for Jordan. Sub-goals: (1) collect metrics from three different systems, (2) analyze trends, (3) draft the report, (4) request review, (5) publish. Describe how you would decompose this into an executable plan, including which tools you would use and how you handle failures at each stage.","Decomposition:
1. Metrics collection: call database_query / terminal / API tools for each system; store results in files.
2. Analysis: use code_interpreter to compute trends, averages, anomalies.
3. Draft: generate structured text (markdown) from analysis results.
4. Review: send via email_client or create a review ticket; wait for feedback.
5. Publish: upload to the designated location (file_system or CMS API).
Failure handling: retry with exponential backoff, fall back to alternative data sources, escalate to human if critical path is blocked, log every step for auditability.","long-horizon,decomposition,agent,robustness"
arb_2545,tool_use,selection,medium,"Available tools: terminal, code_interpreter, browser, file_system.
User goal: find the current weather in Berlin.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: terminal
Approximate call: terminal(query=""find the current weather in Berlin"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2546,logical_reasoning,deduction,easy,All Averys own a book. Riley is a Avery. Does Riley own a book? Explain step by step.,"Yes. From the universal statement 'All Averys own a book' and the particular 'Riley is a Avery', it follows by universal instantiation that Riley owns a book.","deduction,syllogism,logic"
arb_2547,mathematical_reasoning,arithmetic,easy,Start with 46. Subtract 49. Then add 27. What is the final value? Show all steps.,"Step 1: 46 - 49 = -3
Step 2: -3 + 27 = 24
Final answer: 24","math,arithmetic,step-by-step"
arb_2548,tool_use,selection,medium,"Available tools: web_search, terminal, browser.
User goal: calculate the factorial of 12.
Which tool(s) should the agent call first, and with what approximate arguments? Explain the reasoning and the expected observation.","Primary tool: web_search
Approximate call: web_search(code=""import math; print(math.factorial(12))"")
Reasoning: The goal requires information or an action that this tool is designed for. After receiving the observation, the agent should verify completeness and, if needed, call additional tools.","tool-use,function-calling,agent"
arb_2549,agentic_planning,multi_step,medium,You are an agent assisting Cameron. Goal: retrieve the badge from the lab and deliver it to the lobby. Available tools include calendar and file_system. Provide a numbered step-by-step plan with at least 6 steps. Include checks for success and recovery actions.,"1. Use calendar or file_system to locate the current position of the badge (confirm it is in lab).
2. Navigate or request access to the lab.
3. Pick up / retrieve the badge.
4. Verify possession of the badge (success check).
5. Navigate to the lobby.
6. Deliver / place the badge in the lobby and confirm delivery.
Recovery: If step 1 fails, expand search; if navigation fails, request alternative path or human assistance.","planning,agent,multi-step,recovery"
arb_2550,code_reasoning,debugging,medium,"What is wrong with the following Python code?
```python
s = 'hello'
print(s.append('!'))
```
Explain the error and provide a corrected version.","AttributeError: 'str' object has no attribute 'append'. Strings are immutable; use s + '!'.
Corrected version depends on the intended behavior; e.g. fix the operator, index, indentation, key, or method accordingly.","code,debugging,python"