podstack/Shaktimaan
Text Generation • 0.5B • Updated • 17 • 1
id stringlengths 5 34 | source stringclasses 7
values | domain stringclasses 22
values | kind stringclasses 4
values | gold stringlengths 1 4.15k | prompt stringlengths 16 6.51k | has_image bool 1
class | est_tokens int64 6 259k | prompt_truncated bool 2
classes | router_input stringlengths 35 6.53k | route stringclasses 150
values | agent_results stringlengths 588 666 | messages listlengths 3 3 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
math-904 | math | math | math | 1016 | The function $f(n)$ is defined on the positive integers such that $f(f(n)) = 2n$ and $f(4n + 1) = 4n + 3$ for all positive integers $n.$ Find $f(1000).$ | false | 52 | false | [input ~0k tokens]
The function $f(n)$ is defined on the positive integers such that $f(f(n)) = 2n$ and $f(4n + 1) = 4n + 3$ for all positive integers $n.$ Find $f(1000).$ | {"domain": "math", "difficulty": "medium", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 1696, "latency": 8.051527500152588, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7884, "latency": 140.45816159248352, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 458.4477868080139, "skippe... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mbpp-375 | mbpp | code | code |
assert round_num(4722,10)==4720
assert round_num(1111,5)==1110
assert round_num(219,2)==218 | Write a function to round the given number to the nearest multiple of a specific number.
Your code should pass these tests:
assert round_num(4722,10)==4720
assert round_num(1111,5)==1110
assert round_num(219,2)==218 | false | 72 | false | [input ~0k tokens]
Write a function to round the given number to the nearest multiple of a specific number.
Your code should pass these tests:
assert round_num(4722,10)==4720
assert round_num(1111,5)==1110
assert round_num(219,2)==218 | {"domain": "code", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 30, "latency": 0.17873716354370117, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 6654, "latency": 117.27256989479065, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 3396, "latency": 239.34495329856873, "skipped... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-278 | gsm8k | math | num | 1500 | It will take Richard and Sarah 3 years to save enough for a down payment on a house at their current rates. If the house's downpayment is $108000, calculate the amount each person saves per month if they share the monthly savings. | false | 77 | false | [input ~0k tokens]
It will take Richard and Sarah 3 years to save enough for a down payment on a house at their current rates. If the house's downpayment is $108000, calculate the amount each person saves per month if they share the monthly savings. | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 142, "latency": 0.7261126041412354, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 848, "latency": 14.960115432739258, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 1096, "latency": 61.14135789871216, "skipped":... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
humaneval-163 | humaneval | code | code | def check(candidate):
# Check some simple cases
assert candidate(2, 10) == [2, 4, 6, 8], "Test 1"
assert candidate(10, 2) == [2, 4, 6, 8], "Test 2"
assert candidate(132, 2) == [2, 4, 6, 8], "Test 3"
assert candidate(17,89) == [], "Test 4"
# Check some edge cases that are easy to work out by ha... | Complete the following Python function:
```python
def generate_integers(a, b):
"""
Given two positive integers a and b, return the even digits between a
and b, in ascending order.
For example:
generate_integers(2, 8) => [2, 4, 6, 8]
generate_integers(8, 2) => [2, 4, 6, 8]
generate_integers... | false | 116 | false | [input ~0k tokens]
Complete the following Python function:
```python
def generate_integers(a, b):
"""
Given two positive integers a and b, return the even digits between a
and b, in ascending order.
For example:
generate_integers(2, 8) => [2, 4, 6, 8]
generate_integers(8, 2) => [2, 4, 6, 8]
... | {"domain": "code", "difficulty": "medium", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 184, "latency": 4.653826713562012, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7820, "latency": 152.3794822692871, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 728.3827126026154, "skipped"... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mmlupro-995 | mmlu_pro | economics | mc | D | Why do economists make seasonal adjustmentsin economic data before trying to isolate long-term trends?
A. To differentiate between urban and rural economic activity
B. To calculate economic growth
C. To understand consumer behavior
D. To gain a more accurate picture of the actual trend
E. To adjust for variations in w... | false | 184 | false | [input ~0k tokens]
Why do economists make seasonal adjustmentsin economic data before trying to isolate long-term trends?
A. To differentiate between urban and rural economic activity
B. To calculate economic growth
C. To understand consumer behavior
D. To gain a more accurate picture of the actual trend
E. To adjust ... | {"domain": "economics", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 20, "latency": 0.14783143997192383, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 506, "latency": 8.943398714065552, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 448, "latency": 25.00115442276001, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-120 | math | math | math | \frac{2}{3} | The line $y - x \sqrt{3} + 3 = 0$ intersects the parabola $2y^2 = 2x + 3$ at points $A$ and $B.$ Let $P = (\sqrt{3},0).$ Find $|AP - BP|.$ | false | 47 | false | [input ~0k tokens]
The line $y - x \sqrt{3} + 3 = 0$ intersects the parabola $2y^2 = 2x + 3$ at points $A$ and $B.$ Let $P = (\sqrt{3},0).$ Find $|AP - BP|.$ | {"domain": "math", "difficulty": "hard", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 1430, "latency": 6.615198612213135, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7889, "latency": 139.48615980148315, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 5087, "latency": 283.0945107936859, "skippe... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mbpp-565 | mbpp | code | code |
assert split('python') == ['p','y','t','h','o','n']
assert split('Name') == ['N','a','m','e']
assert split('program') == ['p','r','o','g','r','a','m'] | Write a python function to split a string into characters.
Your code should pass these tests:
assert split('python') == ['p','y','t','h','o','n']
assert split('Name') == ['N','a','m','e']
assert split('program') == ['p','r','o','g','r','a','m'] | false | 82 | false | [input ~0k tokens]
Write a python function to split a string into characters.
Your code should pass these tests:
assert split('python') == ['p','y','t','h','o','n']
assert split('Name') == ['N','a','m','e']
assert split('program') == ['p','r','o','g','r','a','m'] | {"domain": "code", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 76, "latency": 0.49846410751342773, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 455, "latency": 8.06350040435791, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 654, "latency": 39.77091932296753, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-390 | gsm8k | math | num | 16 | Fred spent half of his allowance going to the movies. He washed the family car and earned 6 dollars. What is his weekly allowance if he ended with 14 dollars? | false | 53 | false | [input ~0k tokens]
Fred spent half of his allowance going to the movies. He washed the family car and earned 6 dollars. What is his weekly allowance if he ended with 14 dollars? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 131, "latency": 0.6572225093841553, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 362, "latency": 6.420671701431274, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 434, "latency": 24.1644446849823, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-655 | arc | science | mc | C | Which of the following Earth layers has the greatest density?
A. crust
B. mantle
C. inner core
D. outer core | false | 37 | false | [input ~0k tokens]
Which of the following Earth layers has the greatest density?
A. crust
B. mantle
C. inner core
D. outer core | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 10, "latency": 0.13819074630737305, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 306, "latency": 5.494760751724243, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 116, "latency": 6.545129060745239, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-160 | arc | science | mc | A | Which is evidence of a chemical reaction?
A. the light produced by magnesium when burned
B. the evaporation of water from a solution
C. the fizzing of a soft drink
D. the heat from a light bulb | false | 65 | false | [input ~0k tokens]
Which is evidence of a chemical reaction?
A. the light produced by magnesium when burned
B. the evaporation of water from a solution
C. the fizzing of a soft drink
D. the heat from a light bulb | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 16, "latency": 0.13116884231567383, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 764, "latency": 13.699417352676392, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 622, "latency": 34.52627778053284, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-471 | math | math | math | -30 | Evaluate: $-\left(14\div 2\cdot 9-60+3\cdot 9\right)$. | false | 19 | false | [input ~0k tokens]
Evaluate: $-\left(14\div 2\cdot 9-60+3\cdot 9\right)$. | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 6, "latency": 0.09072470664978027, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 957, "latency": 16.88626718521118, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 441, "latency": 24.570704221725464, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-206 | gsm8k | math | num | 54 | Sally picks 30 seashells on Monday. On Tuesday, she picks half as many seashells as she did on Monday. If she can sell each seashell for $1.20, how much money can she make if she sells all of her seashells? | false | 69 | false | [input ~0k tokens]
Sally picks 30 seashells on Monday. On Tuesday, she picks half as many seashells as she did on Monday. If she can sell each seashell for $1.20, how much money can she make if she sells all of her seashells? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 5, "latency": 0.07491135597229004, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 368, "latency": 6.662408351898193, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 456, "latency": 25.73334836959839, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
longbench-66f3edfb821e116aacb2fc91 | longbench | long:Long In-context Learning | mc | C | Read the following text and answer the question.
2023.09
v1.0
User Manual
Searching for Keywords
Search for keywords such as “battery” and “install” to find a topic. If you are using Adobe
Acrobat Reader to read this document, press Ctrl+F on Windows or Command+F on Mac
to begin a search.
Navigating to a Topic
V... | false | 61,062 | true | [input ~61k tokens]
Read the following text and answer the question.
2023.09
v1.0
User Manual
Searching for Keywords
Search for keywords such as “battery” and “install” to find a topic. If you are using Adobe
Acrobat Reader to read this document, press Ctrl+F on Windows or Command+F on Mac
to begin a search.
Nav... | {"domain": "long:Long In-context Learning", "difficulty": "hard", "strategy": "single", "agents": ["nemotron-ultra"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 0, "latency": 0, "skipped": "context"}, "kimi-vl": {"correct": false, "completion_tokens": 0, "latency": 0, "skipped": "context"}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 132.50231194496155, "skipped": null}, "glm-5.3-flash": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mmlupro-112 | mmlu_pro | economics | mc | B | If a monopsony labor market suddenly were transformed into a perfectly competitive labor market, how would the wage and employment change?
A. The wage would decrease, but employment would remain constant.
B. Both would increase.
C. Both would decrease.
D. The wage would remain constant, but employment would increase.
... | false | 217 | false | [input ~0k tokens]
If a monopsony labor market suddenly were transformed into a perfectly competitive labor market, how would the wage and employment change?
A. The wage would decrease, but employment would remain constant.
B. Both would increase.
C. Both would decrease.
D. The wage would remain constant, but employme... | {"domain": "economics", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 14, "latency": 0.12511134147644043, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 738, "latency": 14.352567911148071, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 520, "latency": 29.47224259376526, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-876 | math | math | math | 704 | In an office at various times during the day, the boss gives the secretary a letter to type, each time putting the letter on top of the pile in the secretary's in-box. When there is time, the secretary takes the top letter off the pile and types it. There are nine letters to be typed during the day, and the boss delive... | false | 381 | false | [input ~0k tokens]
In an office at various times during the day, the boss gives the secretary a letter to type, each time putting the letter on top of the pile in the secretary's in-box. When there is time, the secretary takes the top letter off the pile and types it. There are nine letters to be typed during the day, ... | {"domain": "math", "difficulty": "hard", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 135, "latency": 0.7189464569091797, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7555, "latency": 132.23308324813843, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 457.0547134876251, "skippe... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-420 | arc | science | mc | A | Scientists conduct investigations to answer questions. Before making a valid conclusion, scientists must
A. collect relevant evidence
B. tell people about the data
C. publish results from the investigation
D. discuss the investigation with other scientists | false | 86 | false | [input ~0k tokens]
Scientists conduct investigations to answer questions. Before making a valid conclusion, scientists must
A. collect relevant evidence
B. tell people about the data
C. publish results from the investigation
D. discuss the investigation with other scientists | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 11, "latency": 0.09307479858398438, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 466, "latency": 8.915863752365112, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 115, "latency": 6.4524455070495605, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
longbench-66f58d6c821e116aacb33e76 | longbench | long:Single-Document QA | mc | C | Read the following text and answer the question.
Article
iEnergy
Lithium-ion Battery Simulation Optimization and Lifetime
Prediction
Xinqi Xie1, Jun Wangkai1, Weixiang Shen2, Rui Xiong1,*
1 Department of Vehicle Engineering, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, ... | false | 21,058 | true | [input ~21k tokens]
Read the following text and answer the question.
Article
iEnergy
Lithium-ion Battery Simulation Optimization and Lifetime
Prediction
Xinqi Xie1, Jun Wangkai1, Weixiang Shen2, Rui Xiong1,*
1 Department of Vehicle Engineering, School of Mechanical Engineering, Beijing Institute of Technolo... | {"domain": "long:Single-Document QA", "difficulty": "hard", "strategy": "ensemble", "agents": ["glm-5.3-flash", "qwen3.8-27b", "nemotron-ultra"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 11, "latency": 0.326399564743042, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 0, "latency": 0, "skipped": "context"}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 512.8330640792847, "skipped": null}, "glm-5... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-319 | math | math | math | 2 | Solve for $c$: $$\sqrt{4+\sqrt{8+4c}}+ \sqrt{2+\sqrt{2+c}} = 2+2\sqrt{2}$$ | false | 25 | false | [input ~0k tokens]
Solve for $c$: $$\sqrt{4+\sqrt{8+4c}}+ \sqrt{2+\sqrt{2+c}} = 2+2\sqrt{2}$$ | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 1372, "latency": 6.283348083496094, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 4426, "latency": 74.61076235771179, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 841, "latency": 47.34896421432495, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-795 | math | math | math | 8 | At Frank's Fruit Market, 3 bananas cost as much as 2 apples, and 6 apples cost as much as 4 oranges. How many oranges cost as much as 18 bananas? | false | 49 | false | [input ~0k tokens]
At Frank's Fruit Market, 3 bananas cost as much as 2 apples, and 6 apples cost as much as 4 oranges. How many oranges cost as much as 18 bananas? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 1414, "latency": 6.819034576416016, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 2212, "latency": 39.22579288482666, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 235, "latency": 13.176538705825806, "skipped":... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-416 | gsm8k | math | num | 28 | In a zoo, a hare is 14 inches tall, while a camel is 24 times taller than the hare. How tall is the camel in feet? | false | 39 | false | [input ~0k tokens]
In a zoo, a hare is 14 inches tall, while a camel is 24 times taller than the hare. How tall is the camel in feet? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 190, "latency": 0.9727873802185059, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 453, "latency": 8.711654663085938, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 475, "latency": 26.620959520339966, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-509 | gsm8k | math | num | 90 | Iris went to the mall to buy clothes. She bought three jackets at $10 each, two pairs of shorts at $6 each, and four pairs of pants at $12 each. How much did she spend in all? | false | 59 | false | [input ~0k tokens]
Iris went to the mall to buy clothes. She bought three jackets at $10 each, two pairs of shorts at $6 each, and four pairs of pants at $12 each. How much did she spend in all? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 41, "latency": 0.289684534072876, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 663, "latency": 12.497256994247437, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 382, "latency": 21.70706295967102, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-715 | arc | science | mc | A | In the rock cycle, weathering and erosion begin the process of the formation of
A. sediments.
B. magma.
C. minerals.
D. crystals. | false | 44 | false | [input ~0k tokens]
In the rock cycle, weathering and erosion begin the process of the formation of
A. sediments.
B. magma.
C. minerals.
D. crystals. | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 10, "latency": 0.12189865112304688, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 538, "latency": 9.749409198760986, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 138, "latency": 8.093680143356323, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-660 | gsm8k | math | num | 2 | Last week, Charles bought a t-shirt for $30, a backpack for $10, and a blue cap for $5. The store gave him a discount and, in total, he only spent $43. How much of a discount did they give him, in dollars? | false | 69 | false | [input ~0k tokens]
Last week, Charles bought a t-shirt for $30, a backpack for $10, and a blue cap for $5. The store gave him a discount and, in total, he only spent $43. How much of a discount did they give him, in dollars? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 6, "latency": 0.1046152114868164, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 312, "latency": 6.055600881576538, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 288, "latency": 16.222296953201294, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-592 | math | math | math | C | Which of the cones below can be formed from a $252^{\circ}$ sector of a circle of radius 10 by aligning the two straight sides?
[asy]
draw((5.8,8.1)..(-10,0)--(0,0)--(3.1,-9.5)..cycle);
label("10",(-5,0),S);
label("$252^{\circ}$",(0,0),NE);
[/asy]
A. base radius = 6, slant =10
B. base radius = 6, height =10
C. base... | false | 136 | false | [input ~0k tokens]
Which of the cones below can be formed from a $252^{\circ}$ sector of a circle of radius 10 by aligning the two straight sides?
[asy]
draw((5.8,8.1)..(-10,0)--(0,0)--(3.1,-9.5)..cycle);
label("10",(-5,0),S);
label("$252^{\circ}$",(0,0),NE);
[/asy]
A. base radius = 6, slant =10
B. base radius = 6, ... | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 7, "latency": 0.12377309799194336, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 1543, "latency": 26.94293212890625, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 701, "latency": 39.32786703109741, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-145 | math | math | math | 16 | Let $x,$ $y,$ and $z$ be positive real numbers such that $x + y + z = 1.$ Find the minimum value of
\[\frac{x + y}{xyz}.\] | false | 42 | false | [input ~0k tokens]
Let $x,$ $y,$ and $z$ be positive real numbers such that $x + y + z = 1.$ Find the minimum value of
\[\frac{x + y}{xyz}.\] | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 969, "latency": 4.3138837814331055, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7894, "latency": 139.5288429260254, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 767, "latency": 43.264004707336426, "skipped"... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-311 | gsm8k | math | num | 65 | Tom got 40 oranges and 70 apples. If he sold 1/4 of the oranges and 1/2 of the apples. How many fruits were left in total? | false | 41 | false | [input ~0k tokens]
Tom got 40 oranges and 70 apples. If he sold 1/4 of the oranges and 1/2 of the apples. How many fruits were left in total? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 5, "latency": 0.06917643547058105, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 387, "latency": 7.09138560295105, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 290, "latency": 16.406425714492798, "skipped": nu... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-442 | arc | science | mc | D | Which system has layers of smooth muscle tissue that contract to move solid and liquid nutrients and waste through the body?
A. respiratory
B. skeletal
C. endocrine
D. digestive | false | 60 | false | [input ~0k tokens]
Which system has layers of smooth muscle tissue that contract to move solid and liquid nutrients and waste through the body?
A. respiratory
B. skeletal
C. endocrine
D. digestive | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 9, "latency": 0.13938260078430176, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 396, "latency": 8.008780002593994, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 140, "latency": 7.902692079544067, "skipped": nul... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-230 | math | math | math | 32\text{ square cm} | The shaded region consists of 16 congruent squares. If $PQ = 6$ cm, what is the area of the entire shaded region?
[asy]
for(int i = 0; i < 5; ++i)
{
for(int j = 0; j < 2; ++j)
{
filldraw((i,j)--(i+1,j)--(i+1,j+1)--(i,j+1)--(i,j)--cycle,gray,linewidth(2));
}
}
for(int i = 0; i < 2; ++i)
{
for(int j = 0; j < 5; ++... | false | 171 | false | [input ~0k tokens]
The shaded region consists of 16 congruent squares. If $PQ = 6$ cm, what is the area of the entire shaded region?
[asy]
for(int i = 0; i < 5; ++i)
{
for(int j = 0; j < 2; ++j)
{
filldraw((i,j)--(i+1,j)--(i+1,j+1)--(i,j+1)--(i,j)--cycle,gray,linewidth(2));
}
}
for(int i = 0; i < 2; ++i)
{
for(i... | {"domain": "math", "difficulty": "hard", "strategy": "ensemble", "agents": ["glm-5.3-flash", "qwen3.8-27b", "nemotron-ultra"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 785, "latency": 3.592319965362549, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7765, "latency": 137.20805311203003, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 1842, "latency": 102.84183692932129, "skippe... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-76 | math | math | math | 29 | If $f(x) = 3x+2$ and $g(x) = (x-1)^2$, what is $f(g(-2))$? | false | 20 | false | [input ~0k tokens]
If $f(x) = 3x+2$ and $g(x) = (x-1)^2$, what is $f(g(-2))$? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 121, "latency": 0.7788875102996826, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 508, "latency": 9.137727499008179, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 346, "latency": 19.511282682418823, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-618 | gsm8k | math | num | 4 | Rose bought her mother one dozen flowers. Two of the flowers are daisies. Three-fifths of the remaining flowers are tulips and the rest are sunflowers. How many sunflowers are there? | false | 61 | false | [input ~0k tokens]
Rose bought her mother one dozen flowers. Two of the flowers are daisies. Three-fifths of the remaining flowers are tulips and the rest are sunflowers. How many sunflowers are there? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 246, "latency": 1.1356141567230225, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 345, "latency": 6.939244031906128, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 260, "latency": 14.616687774658203, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-408 | math | math | math | 4 | $ABCDE$ is a regular pentagon. $AP$, $AQ$ and $AR$ are the perpendiculars dropped from $A$ onto $CD$, $CB$ extended and $DE$ extended, respectively. Let $O$ be the center of the pentagon. If $OP = 1$, then find $AO + AQ + AR$.
[asy]
unitsize(2 cm);
pair A, B, C, D, E, O, P, Q, R;
A = dir(90);
B = dir(90 - 360/5... | false | 274 | false | [input ~0k tokens]
$ABCDE$ is a regular pentagon. $AP$, $AQ$ and $AR$ are the perpendiculars dropped from $A$ onto $CD$, $CB$ extended and $DE$ extended, respectively. Let $O$ be the center of the pentagon. If $OP = 1$, then find $AO + AQ + AR$.
[asy]
unitsize(2 cm);
pair A, B, C, D, E, O, P, Q, R;
A = dir(90);
... | {"domain": "math", "difficulty": "medium", "strategy": "single", "agents": ["qwen3.8-27b"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 1213, "latency": 5.264846324920654, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7662, "latency": 133.40829062461853, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 323.8845920562744, "skippe... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mbpp-80 | mbpp | code | code |
assert tetrahedral_number(5) == 35.0
assert tetrahedral_number(6) == 56.0
assert tetrahedral_number(7) == 84.0 | Write a function to find the nth tetrahedral number.
Your code should pass these tests:
assert tetrahedral_number(5) == 35.0
assert tetrahedral_number(6) == 56.0
assert tetrahedral_number(7) == 84.0 | false | 67 | false | [input ~0k tokens]
Write a function to find the nth tetrahedral number.
Your code should pass these tests:
assert tetrahedral_number(5) == 35.0
assert tetrahedral_number(6) == 56.0
assert tetrahedral_number(7) == 84.0 | {"domain": "code", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 143, "latency": 1.025465726852417, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 1135, "latency": 20.25847864151001, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 808, "latency": 55.788366079330444, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-169 | math | math | math | 4970 | How many four-digit numbers greater than 2999 can be formed such that the product of the middle two digits exceeds 5? | false | 40 | false | [input ~0k tokens]
How many four-digit numbers greater than 2999 can be formed such that the product of the middle two digits exceeds 5? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 975, "latency": 4.442337274551392, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 4225, "latency": 74.47156190872192, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 2268, "latency": 126.2984721660614, "skipped":... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-671 | arc | science | mc | D | Which food provides the most energy for the body in the shortest amount of time?
A. potato
B. meat
C. milk
D. fruit | false | 39 | false | [input ~0k tokens]
Which food provides the most energy for the body in the shortest amount of time?
A. potato
B. meat
C. milk
D. fruit | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 9, "latency": 0.12296843528747559, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 490, "latency": 8.964405059814453, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 181, "latency": 10.306198596954346, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mbpp-882 | mbpp | code | code |
assert parallelogram_perimeter(10,20)==400
assert parallelogram_perimeter(15,20)==600
assert parallelogram_perimeter(8,9)==144 | Write a function to caluclate perimeter of a parallelogram.
Your code should pass these tests:
assert parallelogram_perimeter(10,20)==400
assert parallelogram_perimeter(15,20)==600
assert parallelogram_perimeter(8,9)==144 | false | 74 | false | [input ~0k tokens]
Write a function to caluclate perimeter of a parallelogram.
Your code should pass these tests:
assert parallelogram_perimeter(10,20)==400
assert parallelogram_perimeter(15,20)==600
assert parallelogram_perimeter(8,9)==144 | {"domain": "code", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 85, "latency": 0.5087344646453857, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7862, "latency": 136.77135181427002, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 1756, "latency": 170.3317449092865, "skipped"... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-732 | arc | science | mc | A | Which mechanism is not an example of a negative feedback mechanism in the human body?
A. contracting muscles
B. regulating blood pressure
C. regulating body temperature
D. maintaining blood sugar level | false | 68 | false | [input ~0k tokens]
Which mechanism is not an example of a negative feedback mechanism in the human body?
A. contracting muscles
B. regulating blood pressure
C. regulating body temperature
D. maintaining blood sugar level | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 13, "latency": 0.13135671615600586, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 1038, "latency": 18.368236541748047, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 711, "latency": 39.94104027748108, "skipped":... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-568 | arc | science | mc | A | For a class project, students converted the daily temperatures for the past month from Fahrenheit to Celsius. What is the clearest way for the students to present the information?
A. table
B. formula
C. pie chart
D. line graph | false | 76 | false | [input ~0k tokens]
For a class project, students converted the daily temperatures for the past month from Fahrenheit to Celsius. What is the clearest way for the students to present the information?
A. table
B. formula
C. pie chart
D. line graph | {"domain": "science", "difficulty": "medium", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 9, "latency": 0.11081290245056152, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 727, "latency": 13.934837341308594, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 234, "latency": 13.313121557235718, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
arc-126 | arc | science | mc | D | Annette is using many tools to build a doghouse. For which of the following would Annette most likely use a tape measure while building the doghouse?
A. deciding which type of wood to use
B. removing extra nails from the wood
C. attaching different pieces of wood together
D. determining where to cut the wood into piec... | false | 108 | false | [input ~0k tokens]
Annette is using many tools to build a doghouse. For which of the following would Annette most likely use a tape measure while building the doghouse?
A. deciding which type of wood to use
B. removing extra nails from the wood
C. attaching different pieces of wood together
D. determining where to cut... | {"domain": "science", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 8, "latency": 0.08851146697998047, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 1070, "latency": 18.88690733909607, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 180, "latency": 10.136931896209717, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-431 | math | math | math | \frac{\pi}{8} | A projectile is fired with an initial velocity of $v$ at an angle of $\theta$ from the ground. Then its trajectory can modeled by the parametric equations
\begin{align*}
x &= vt \cos \theta, \\
y &= vt \sin \theta - \frac{1}{2} gt^2,
\end{align*}where $t$ denotes time and $g$ denotes acceleration due to gravity, formi... | false | 505 | false | [input ~0k tokens]
A projectile is fired with an initial velocity of $v$ at an angle of $\theta$ from the ground. Then its trajectory can modeled by the parametric equations
\begin{align*}
x &= vt \cos \theta, \\
y &= vt \sin \theta - \frac{1}{2} gt^2,
\end{align*}where $t$ denotes time and $g$ denotes acceleration du... | {"domain": "math", "difficulty": "medium", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 1192, "latency": 5.517663478851318, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 7431, "latency": 129.9392011165619, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 456.5765256881714, "skipped... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-860 | math | math | math | 1 | Done with her new problems, Wendy takes a break from math. Still without any fresh reading material, she feels a bit antsy. She starts to feel annoyed that Michael's loose papers clutter the family van. Several of them are ripped, and bits of paper litter the floor. Tired of trying to get Michael to clean up after hims... | false | 398 | false | [input ~0k tokens]
Done with her new problems, Wendy takes a break from math. Still without any fresh reading material, she feels a bit antsy. She starts to feel annoyed that Michael's loose papers clutter the family van. Several of them are ripped, and bits of paper litter the floor. Tired of trying to get Michael to ... | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["nemotron-lightning"]} | {"gemma-3-1b": {"correct": true, "completion_tokens": 176, "latency": 0.8449013233184814, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 2900, "latency": 50.98478555679321, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 5239, "latency": 293.2378387451172, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
longbench-66f39ac5821e116aacb2da81 | longbench | long:Code Repository Understanding | mc | B | Read the following text and answer the question.
cmake_minimum_required(VERSION 3.18)
#
# Allow for MSVC Runtime library controls
#
if(POLICY CMP0091)
cmake_policy(SET CMP0091 NEW)
endif()
#
# We use simple syntax in cmake_dependent_option, so we are compatible with the
# extended syntax in CMake 3.22+
# https://c... | false | 115,087 | true | [input ~115k tokens]
Read the following text and answer the question.
cmake_minimum_required(VERSION 3.18)
#
# Allow for MSVC Runtime library controls
#
if(POLICY CMP0091)
cmake_policy(SET CMP0091 NEW)
endif()
#
# We use simple syntax in cmake_dependent_option, so we are compatible with the
# extended syntax in CM... | {"domain": "long:Code Repository Understanding", "difficulty": "hard", "strategy": "single", "agents": ["nemotron-ultra"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 0, "latency": 0, "skipped": "context"}, "kimi-vl": {"correct": false, "completion_tokens": 0, "latency": 0, "skipped": "context"}, "nemotron-lightning": {"correct": false, "completion_tokens": 2009, "latency": 46.06229853630066, "skipped": null}, "glm-5.3-flash": {... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
gsm8k-76 | gsm8k | math | num | 252 | Juanita goes through 1 bottle of sunscreen a month. She likes to stock up for the entire year when a certain website offers 30% off her entire purchase. If each bottle is $30.00 how much will all of the sunscreen cost after the discount? | false | 80 | false | [input ~0k tokens]
Juanita goes through 1 bottle of sunscreen a month. She likes to stock up for the entire year when a certain website offers 30% off her entire purchase. If each bottle is $30.00 how much will all of the sunscreen cost after the discount? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["kimi-vl"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 5, "latency": 0.08140397071838379, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 504, "latency": 8.833036422729492, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 539, "latency": 30.318421602249146, "skipped": n... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-824 | math | math | math | 56 | Consider all polynomials of the form
\[x^9 + a_8 x^8 + a_7 x^7 + \dots + a_2 x^2 + a_1 x + a_0,\]where $a_i \in \{0,1\}$ for all $0 \le i \le 8.$ Find the number of such polynomials that have exactly two different integer roots. | false | 77 | false | [input ~0k tokens]
Consider all polynomials of the form
\[x^9 + a_8 x^8 + a_7 x^7 + \dots + a_2 x^2 + a_1 x + a_0,\]where $a_i \in \{0,1\}$ for all $0 \le i \le 8.$ Find the number of such polynomials that have exactly two different integer roots. | {"domain": "math", "difficulty": "medium", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 773, "latency": 3.7892961502075195, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 4488, "latency": 80.00131583213806, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 8192, "latency": 460.4151380062103, "skipped... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
mmlupro-205 | mmlu_pro | economics | mc | A | An American buys an entertainment system that was manufactured in China. How does the U.S. national income accounts treat this transaction?
A. Net exports and GDP both fall.
B. GDP falls and there is no change in net exports.
C. Net exports rise and there is no change in GDP.
D. Net exports and GDP both rise.
E. Net e... | false | 196 | false | [input ~0k tokens]
An American buys an entertainment system that was manufactured in China. How does the U.S. national income accounts treat this transaction?
A. Net exports and GDP both fall.
B. GDP falls and there is no change in net exports.
C. Net exports rise and there is no change in GDP.
D. Net exports and GDP ... | {"domain": "economics", "difficulty": "hard", "strategy": "ensemble", "agents": ["glm-5.3-flash", "qwen3.8-27b", "nemotron-ultra"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 17, "latency": 0.14119911193847656, "skipped": null}, "kimi-vl": {"correct": false, "completion_tokens": 645, "latency": 12.337969064712524, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 918, "latency": 51.24924349784851, "skipped"... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-642 | math | math | math | \left( 0, -\frac{47}{16} \right) | Find the focus of the parabola $y = 4x^2 - 3.$ | false | 16 | false | [input ~0k tokens]
Find the focus of the parabola $y = 4x^2 - 3.$ | {"domain": "math", "difficulty": "hard", "strategy": "ensemble", "agents": ["glm-5.3-flash", "qwen3.8-27b", "nemotron-ultra"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 90, "latency": 0.458341121673584, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 1911, "latency": 33.845388412475586, "skipped": null}, "nemotron-lightning": {"correct": false, "completion_tokens": 626, "latency": 35.22683095932007, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
math-537 | math | math | math | 50\% | Eighty percent of adults drink coffee and seventy percent drink tea. What is the smallest possible percent of adults who drink both coffee and tea? | false | 50 | false | [input ~0k tokens]
Eighty percent of adults drink coffee and seventy percent drink tea. What is the smallest possible percent of adults who drink both coffee and tea? | {"domain": "math", "difficulty": "easy", "strategy": "single", "agents": ["glm-5.3-flash"]} | {"gemma-3-1b": {"correct": false, "completion_tokens": 254, "latency": 1.136509895324707, "skipped": null}, "kimi-vl": {"correct": true, "completion_tokens": 1484, "latency": 25.968236684799194, "skipped": null}, "nemotron-lightning": {"correct": true, "completion_tokens": 428, "latency": 23.81358504295349, "skipped": ... | [
{
"role": "system",
"content": "You are Shaktimaan, an orchestrator. You do not answer the task yourself. You read the task and decide which agent(s) from the pool should handle it, minimising cost while maximising the chance of a correct answer.\nAgents (cheapest first): gemma-3-1b, kimi-vl, nemotron-light... |
End of preview. Expand in Data Studio
Shaktimaan Routing Data v1.0
The training and evaluation data behind Shaktimaan v1.0, a 0.5B LLM router. Each row is one task that every capable agent in a six-model pool attempted, with every answer automatically graded, plus the cost-aware routing label derived from those outcomes. Useful for training or benchmarking LLM routers, model-selection policies and cascades.
Published: 4720 train / 523 test rows (5243 tasks). The model was additionally trained and evaluated on 300 ChartQA rows that are not published here (see Licensing).
Agent pool
| agent | model | context | image input (in this pool) | relative cost / token |
|---|---|---|---|---|
gemma-3-1b |
google/gemma-3-1b-it |
32k | no | 0.02 |
kimi-vl |
Kimi-VL-A3B-Thinking |
8k | yes | 0.05 |
nemotron-lightning |
nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 |
128k | no | 0.1 |
glm-5.3-flash |
zai-org/GLM-5.3-Flash |
32k | no | 0.2 |
qwen3.8-27b |
Qwen/Qwen3.8-27B |
32k | no | 0.5 |
nemotron-ultra |
nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4 |
256k | no | 1.0 |
"Relative cost" is a per-token price weight used to compute routing cost; it reflects each model's relative serving cost (roughly active parameters), not a vendor price.
Sources
| source | category | train | test | grading | license |
|---|---|---|---|---|---|
| GSM8K | math word problems | 895 | 105 | numeric match | MIT |
| MATH | competition math | 894 | 106 | normalised LaTeX / numeric match | MIT |
| MMLU-Pro | knowledge MCQ (10 options) | 1080 | 120 | option letter | MIT |
| ARC-Challenge | science MCQ | 725 | 75 | option letter | CC-BY-SA-4.0 |
| MBPP | Python coding | 633 | 67 | unit tests executed in a network-less sandbox | CC-BY-4.0 |
| HumanEval | Python coding | 146 | 18 | unit tests executed in a network-less sandbox | MIT |
| LongBench-v2 | long-context MCQ (up to ~260k tokens) | 347 | 32 | option letter | Apache-2.0 |
| ChartQA | chart images | 269 | 31 | relaxed match (numbers within 5%) | GPL-3.0 (not published here) |
Long LongBench-v2 contexts are stored truncated to what the router reads (first 5,000 and last 1,500 characters);
est_tokens keeps the original size.
How it was built
- Solve - every agent answered every task once (temperature 0.6, up to 8,192 output tokens). Agents that cannot
take an input (no image support, context window too small) are recorded as failures without being called
(
skipped:no_vision/context). - Grade - automatically, per source (see table). Code is graded by running the dataset's unit tests against the model's code in a throwaway container with no network, 512 MB RAM and a 15 s limit.
- Label - the cheapest agent that solved the task, counting a win only if that agent is reliable on the source
(solve rate >= 70%), because single-sample wins by weak agents are mostly luck a router cannot predict.
gemma-3-1bis never a label. If no eligible agent solved the task, the label is an ensemble of the capable strong agents (glm-5.3-flash,qwen3.8-27b,nemotron-ultra, aggregated by the last), or the strongest capable agent alone when fewer than two can take the input.
Agent solve rates (single attempt, all 5543 tasks incl. ChartQA)
| source | gemma-3-1b |
kimi-vl |
nemotron-lightning |
glm-5.3-flash |
qwen3.8-27b |
nemotron-ultra |
|---|---|---|---|---|---|---|
| GSM8K | 26.3% | 78.0% | 95.0% | 96.9% | 96.8% | 95.6% |
| MATH | 18.4% | 41.9% | 70.7% | 75.8% | 75.9% | 71.6% |
| MMLU-Pro | 12.7% | 50.1% | 73.0% | 81.0% | 82.5% | 76.2% |
| ARC-Challenge | 44.9% | 91.4% | 97.9% | 97.8% | 98.0% | 98.2% |
| MBPP | 50.4% | 67.0% | 90.9% | 94.0% | 88.4% | 93.6% |
| HumanEval | 41.5% | 79.3% | 92.7% | 92.1% | 93.3% | 97.0% |
| LongBench-v2 | 5.3% | 0.0% | 25.9% | 10.3% | 12.1% | 54.6% |
| ChartQA | 0.0% | 80.0% | 0.0% | 0.0% | 0.0% | 0.0% |
| all | 25.2% | 60.8% | 75.8% | 78.1% | 77.9% | 79.3% |
Label distribution (all 5543 tasks incl. ChartQA)
| route label | tasks | share |
|---|---|---|
glm-5.3-flash |
1854 | 33.4% |
nemotron-lightning |
1509 | 27.2% |
kimi-vl |
1373 | 24.8% |
glm-5.3-flash,qwen3.8-27b,nemotron-ultra (ensemble) |
365 | 6.6% |
nemotron-ultra |
329 | 5.9% |
qwen3.8-27b |
113 | 2.0% |
Fields
| field | description |
|---|---|
id, source, domain, kind |
task id, source dataset, domain, grading kind (num, math, mc, code, relaxed) |
prompt |
the task (LongBench contexts truncated, see above) |
gold |
reference answer; for code, the unit tests |
has_image, est_tokens, prompt_truncated |
input metadata (est_tokens is a conservative chars/3 estimate) |
router_input |
exactly what the router reads: a size / image header plus the start and end of the task |
agent_results |
JSON: per agent correct, completion_tokens, latency (s), skipped |
route |
JSON label: domain, difficulty, strategy (single / ensemble), agents |
messages |
chat-format SFT example (system, user = router_input, assistant = route) |
Licensing
Released under CC-BY-SA-4.0, as required by the ARC-Challenge share-alike terms; the other included sources are
MIT, CC-BY-4.0 and Apache-2.0. The 300 ChartQA rows used for training are excluded because ChartQA is GPL-3.0;
gen.py in the model repository regenerates them. Please also cite the original datasets.
- Downloads last month
- 32
Models trained or fine-tuned on podstack/Shaktimaan-routing