Add files using upload-large-folder tool

tasks/0000_531_531577_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- Pokemon.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the highest total stat value observed in the Pokémon dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0000_531_531577_qa_1/task.toml

@@ -0,0 +1,57 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0000_531_531577_qa_1"
+description = "What is the highest total stat value observed in the Pokémon dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0000/531/531577.ipynb_qa_1"
+kaggle_dataset_name = "abcsds/pokemon"
+gold_answer = "800"
+reward_mode_initial = "numeric"
+package_tier = 1
+difficulty_level = 1
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 2
+memory_mb = 4096
+storage_mb = 10240
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "abcsds__pokemon"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "abcsds/pokemon"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "780"
+QUESTION = "What is the highest total stat value observed in the Pokémon dataset?"
+REWARD_MODE = "numeric"
+
+[agent]
+timeout_sec = 900.0
+
+[solution.env]

tasks/0000_621_621324_qa_4/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- database.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the most common weapon carried by victims with signs of mental illness in fatal shootings?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0000_621_621324_qa_4/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0000_621_621324_qa_4"
+description = "What is the most common weapon carried by victims with signs of mental illness in fatal shootings?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0000/621/621324.ipynb_qa_4"
+kaggle_dataset_name = "washingtonpost/police-shootings"
+gold_answer = "gun"
+reward_mode_initial = "exact_short"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "washingtonpost__police-shootings"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "washingtonpost/police-shootings"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "gun"
+QUESTION = "What is the most common weapon carried by victims with signs of mental illness in fatal shootings?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0000_806_806826_qa_5/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- archive.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Is the distribution of temperature deviations between February and March approximately normal based on the kernel density estimate and observed frequency counts?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0000_806_806826_qa_5/task.toml

@@ -0,0 +1,56 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0000_806_806826_qa_5"
+description = "Is the distribution of temperature deviations between February and March approximately normal based on the kernel density estimate and observed frequency counts?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0000/806/806826.ipynb_qa_5"
+kaggle_dataset_name = "groundhogclub/groundhog-day"
+gold_answer = "Yes"
+reward_mode_initial = "exact_bool"
+package_tier = 1
+difficulty_level = 4
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 2
+memory_mb = 4096
+storage_mb = 10240
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "groundhogclub__groundhog-day"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "groundhogclub/groundhog-day"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "Yes"
+QUESTION = "Is the distribution of temperature deviations between February and March approximately normal based on the kernel density estimate and observed frequency counts?"
+REWARD_MODE = "exact_bool"
+
+[agent]
+timeout_sec = 900.0
+
+[solution.env]

tasks/0000_965_965367_qa_1/instruction.md

@@ -0,0 +1,16 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- salary.csv
+- team.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+How many World Series winners since 1985 had the highest payroll rank (rank 1)?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0000_965_965367_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0000_965_965367_qa_1"
+description = "How many World Series winners since 1985 had the highest payroll rank (rank 1)?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0000/965/965367.ipynb_qa_1"
+kaggle_dataset_name = "seanlahman/the-history-of-baseball"
+gold_answer = "5"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "seanlahman__the-history-of-baseball"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "seanlahman/the-history-of-baseball"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "5"
+QUESTION = "How many World Series winners since 1985 had the highest payroll rank (rank 1)?"
+REWARD_MODE = "numeric"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0001_068_1068916_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the average spam score across all entries in the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0001_068_1068916_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0001_068_1068916_qa_1"
+description = "What is the average spam score across all entries in the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0001/068/1068916.ipynb_qa_1"
+kaggle_dataset_name = "mrisdal/fake-news"
+gold_answer = "0.026"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "mrisdal__fake-news"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "mrisdal/fake-news"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "0.026"
+QUESTION = "What is the average spam score across all entries in the dataset?"
+REWARD_MODE = "numeric"
+ATOL = "0.001"
+RTOL = "0.005"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0001_085_1085629_qa_2/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What are the top three features according to the feature importance ranking provided by the Extra Trees Classifier?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0001_085_1085629_qa_2/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0001_085_1085629_qa_2"
+description = "What are the top three features according to the feature importance ranking provided by the Extra Trees Classifier?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0001/085/1085629.ipynb_qa_2"
+kaggle_dataset_name = "uciml/adult-census-income"
+gold_answer = "fnlwgt, age, hours.per.week"
+reward_mode_initial = "list"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "uciml__adult-census-income"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "uciml/adult-census-income"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "fnlwgt, age, hours.per.week"
+QUESTION = "What are the top three features according to the feature importance ranking provided by the Extra Trees Classifier?"
+REWARD_MODE = "list"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0001_250_1250662_qa_3/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- survey_results_public.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the average ideal workday start time (in 24-hour format) for R users in the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0001_250_1250662_qa_3/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0001_250_1250662_qa_3"
+description = "What is the average ideal workday start time (in 24-hour format) for R users in the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0001/250/1250662.ipynb_qa_3"
+kaggle_dataset_name = "stackoverflow/so-survey-2017"
+gold_answer = "9.293624"
+reward_mode_initial = "numeric"
+package_tier = 1
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "stackoverflow__so-survey-2017"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "stackoverflow/so-survey-2017"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "9.293624"
+QUESTION = "What is the average ideal workday start time (as a decimal number in 24-hour format, e.g. 9.5 means 9:30 AM) for R users in the dataset?"
+REWARD_MODE = "numeric"
+ATOL = "0.01"
+RTOL = "0.01"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0002_240_2240465_qa_4/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the 90th percentile of USD pledged amount for technology category projects?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0002_240_2240465_qa_4/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0002_240_2240465_qa_4"
+description = "What is the 90th percentile of USD pledged amount for technology category projects?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0002/240/2240465.ipynb_qa_4"
+kaggle_dataset_name = "kemical/kickstarter-projects"
+gold_answer = "21223.50"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "kemical__kickstarter-projects"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "kemical/kickstarter-projects"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "21223.50"
+QUESTION = "What is the 90th percentile of USD pledged amount for technology category projects?"
+REWARD_MODE = "numeric"
+ATOL = "0.05"
+RTOL = "0.01"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0011_222_11222401_qa_5/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- Reviews.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the top negative word contributing to negative sentiment predictions in the model?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0011_222_11222401_qa_5/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0011_222_11222401_qa_5"
+description = "What is the top negative word contributing to negative sentiment predictions in the model?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0011/222/11222401.ipynb_qa_5"
+kaggle_dataset_name = "snap/amazon-fine-food-reviews"
+gold_answer = "disappointing"
+reward_mode_initial = "exact_short"
+package_tier = 1
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "snap__amazon-fine-food-reviews"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "snap/amazon-fine-food-reviews"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "worst"
+QUESTION = "What is the top negative word contributing to negative sentiment predictions in the model?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0016_558_16558367_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Which user has the highest total positive sentiment score according to the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0016_558_16558367_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0016_558_16558367_qa_1"
+description = "Which user has the highest total positive sentiment score according to the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0016/558/16558367.ipynb_qa_1"
+kaggle_dataset_name = "arathee2/demonetization-in-india-twitter-data"
+gold_answer = "mituamin"
+reward_mode_initial = "exact_short"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "arathee2__demonetization-in-india-twitter-data"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "arathee2/demonetization-in-india-twitter-data"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "mituamin"
+QUESTION = "Which user has the highest total positive sentiment score according to the dataset?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0016_734_16734421_qa_3/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Which student performance class (H/M/L) achieved the highest F1-score in the Random Forest model evaluation?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0016_734_16734421_qa_3/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0016_734_16734421_qa_3"
+description = "Which student performance class (H/M/L) achieved the highest F1-score in the Random Forest model evaluation?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0016/734/16734421.ipynb_qa_3"
+kaggle_dataset_name = "aljarah/xAPI-Edu-Data"
+gold_answer = "L"
+reward_mode_initial = "exact_short"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "aljarah__xAPI-Edu-Data"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "aljarah/xAPI-Edu-Data"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "L"
+QUESTION = "Which student performance class (H/M/L) achieved the highest F1-score in the Random Forest model evaluation?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0017_740_17740333_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the correlation coefficient between average monthly hours worked and time spent at the company in the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0017_740_17740333_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0017_740_17740333_qa_1"
+description = "What is the correlation coefficient between average monthly hours worked and time spent at the company in the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0017/740/17740333.ipynb_qa_1"
+kaggle_dataset_name = "jiangzuo/hr-comma-sep"
+gold_answer = "0.12775491036185835"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "jiangzuo__hr-comma-sep"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "jiangzuo/hr-comma-sep"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "0.12775491036185835"
+QUESTION = "What is the correlation coefficient between average monthly hours worked and time spent at the company in the dataset?"
+REWARD_MODE = "flexible"
+ATOL = "0.001"
+RTOL = "0.005"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0020_991_20991108_qa_5/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- sensor_readings_24.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the total number of test samples evaluated in the MLP model's confusion matrix output?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0020_991_20991108_qa_5/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0020_991_20991108_qa_5"
+description = "What is the total number of test samples evaluated in the MLP model's confusion matrix output?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0020/991/20991108.ipynb_qa_5"
+kaggle_dataset_name = "uciml/wall-following-robot"
+gold_answer = "1637 samples"
+reward_mode_initial = "exact_short"
+package_tier = 1
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "uciml__wall-following-robot"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "uciml/wall-following-robot"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "1637 samples"
+QUESTION = "What is the total number of test samples evaluated in the MLP model's confusion matrix output? (Use a 70/30 train/test split with test_size=0.3)"
+REWARD_MODE = "flexible"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0021_751_21751462_qa_4/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- kc_house_data.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+How many duplicate property records were identified and removed based on the unique combination of location, date, and property characteristics?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0021_751_21751462_qa_4/task.toml

@@ -0,0 +1,57 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0021_751_21751462_qa_4"
+description = "How many duplicate property records were identified and removed based on the unique combination of location, date, and property characteristics?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0021/751/21751462.ipynb_qa_4"
+kaggle_dataset_name = "harlfoxem/housesalesprediction"
+gold_answer = "2"
+reward_mode_initial = "numeric"
+package_tier = 1
+difficulty_level = 2
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 2
+memory_mb = 4096
+storage_mb = 10240
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "harlfoxem__housesalesprediction"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "harlfoxem/housesalesprediction"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "2"
+QUESTION = "How many duplicate property records were identified and removed based on the unique combination of location, date, and property characteristics? (Count all rows that are part of any duplicate group, i.e., use keep=False when identifying duplicates.)"
+REWARD_MODE = "numeric"
+
+[agent]
+timeout_sec = 900.0
+
+[solution.env]

tasks/0021_803_21803688_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- diamonds.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Which two features are identified as the most important predictors of diamond price according to the RandomForestRegressor model based on feature importance scores?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0021_803_21803688_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0021_803_21803688_qa_1"
+description = "Which two features are identified as the most important predictors of diamond price according to the RandomForestRegressor model based on feature importance scores?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0021/803/21803688.ipynb_qa_1"
+kaggle_dataset_name = "shivam2503/diamonds"
+gold_answer = "carat and y"
+reward_mode_initial = "exact_short"
+package_tier = 2
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "shivam2503__diamonds"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "shivam2503/diamonds"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "carat and y"
+QUESTION = "Which two features are identified as the most important predictors of diamond price according to the RandomForestRegressor model based on feature importance scores?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0023_721_23721314_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the highest single-season passing yardage achieved by a player in the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0023_721_23721314_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0023_721_23721314_qa_1"
+description = "What is the highest single-season passing yardage achieved by a player in the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0023/721/23721314.ipynb_qa_1"
+kaggle_dataset_name = "jerrinv/nfl-offense-cleaned-2017to2007"
+gold_answer = "5477"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "jerrinv__nfl-offense-cleaned-2017to2007"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "jerrinv/nfl-offense-cleaned-2017to2007"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "5477"
+QUESTION = "What is the highest single-season passing yardage achieved by a player in the dataset?"
+REWARD_MODE = "numeric"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0023_846_23846255_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the absolute difference between the maximum purchase amount and the 75th percentile purchase amount in the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0023_908_23908999_qa_2/task.toml

@@ -0,0 +1,57 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0023_908_23908999_qa_2"
+description = "What is the most frequently used opening move in the dataset, and how many times does it appear?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0023/908/23908999.ipynb_qa_2"
+kaggle_dataset_name = "datasnaek/chess"
+gold_answer = "e4 with 12598 occurrences"
+reward_mode_initial = "exact_short"
+package_tier = 1
+difficulty_level = 2
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 2
+memory_mb = 4096
+storage_mb = 10240
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "datasnaek__chess"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "datasnaek/chess"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "e4 with 12598 occurrences"
+QUESTION = "What is the most frequently used opening move in the dataset, and how many times does it appear?"
+REWARD_MODE = "exact_short"
+
+[agent]
+timeout_sec = 900.0
+
+[solution.env]

tasks/0024_991_24991983_qa_2/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the highest correlation coefficient between any two features in the dataset according to the heatmap visualization?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0024_991_24991983_qa_2/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0024_991_24991983_qa_2"
+description = "What is the highest correlation coefficient between any two features in the dataset according to the heatmap visualization?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0024/991/24991983.ipynb_qa_2"
+kaggle_dataset_name = "yuqing01/breast-cancer"
+gold_answer = "0.98"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "yuqing01__breast-cancer"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "yuqing01/breast-cancer"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "0.98"
+QUESTION = "What is the highest correlation coefficient between any two features in the dataset according to the heatmap visualization?"
+REWARD_MODE = "numeric"
+ATOL = "0.05"
+RTOL = "0.01"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0026_141_26141545_qa_4/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- haberman.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Which combination of variables demonstrated the strongest predictive power for survival outcomes in the bivariate analysis?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0026_141_26141545_qa_4/task.toml

@@ -0,0 +1,57 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0026_141_26141545_qa_4"
+description = "Which combination of variables demonstrated the strongest predictive power for survival outcomes in the bivariate analysis?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0026/141/26141545.ipynb_qa_4"
+kaggle_dataset_name = "gilsousa/habermans-survival-data-set"
+gold_answer = "Age and No.of.AxillaryNodes"
+reward_mode_initial = "flexible"
+package_tier = 1
+difficulty_level = 4
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 2
+memory_mb = 4096
+storage_mb = 10240
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "gilsousa__habermans-survival-data-set"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "gilsousa/habermans-survival-data-set"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "Age and No.of.AxillaryNodes"
+QUESTION = "Which combination of variables demonstrated the strongest predictive power for survival outcomes in the bivariate analysis? (Consider KDE/distribution-based separation between survival classes, not logistic regression AUC. The answer should be a pair of the three feature variables: Age, Year of operation, and No.of.AxillaryNodes.)"
+REWARD_MODE = "flexible"
+
+[agent]
+timeout_sec = 900.0
+
+[solution.env]

tasks/0029_908_29908654_qa_2/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Among the selected samples (43, 12, 39), which one has the highest Grocery spending value?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0029_908_29908654_qa_2/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0029_908_29908654_qa_2"
+description = "Among the selected samples (43, 12, 39), which one has the highest Grocery spending value?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0029/908/29908654.ipynb_qa_2"
+kaggle_dataset_name = "nguyen17/uci-wholesale-customers-data"
+gold_answer = "43"
+reward_mode_initial = "exact_short"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "nguyen17__uci-wholesale-customers-data"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "nguyen17/uci-wholesale-customers-data"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "43"
+QUESTION = "Among the selected samples (43, 12, 39), which one has the highest Grocery spending value?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0032_680_32680563_qa_2/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the total number of training samples for the author with the highest representation in the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0032_680_32680563_qa_2/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0032_680_32680563_qa_2"
+description = "What is the total number of training samples for the author with the highest representation in the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0032/680/32680563.ipynb_qa_2"
+kaggle_dataset_name = "lyuyanhan/spookyauthordata"
+gold_answer = "7900"
+reward_mode_initial = "numeric"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "lyuyanhan__spookyauthordata"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "lyuyanhan/spookyauthordata"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "7900"
+QUESTION = "What is the total number of training samples for the author with the highest representation in the dataset?"
+REWARD_MODE = "numeric"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0034_916_34916115_qa_2/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+Which country produces wines with the highest average price according to the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0034_916_34916115_qa_2/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0034_916_34916115_qa_2"
+description = "Which country produces wines with the highest average price according to the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0034/916/34916115.ipynb_qa_2"
+kaggle_dataset_name = "zynicide/wine-reviews"
+gold_answer = "Switzerland"
+reward_mode_initial = "exact_short"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "zynicide__wine-reviews"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "zynicide/wine-reviews"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "Switzerland"
+QUESTION = "Which country produces wines with the highest average price according to the dataset?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0035_835_35835284_qa_2/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- SolarPrediction.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the hour of the day with the highest average solar radiation based on the dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0035_835_35835284_qa_2/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0035_835_35835284_qa_2"
+description = "What is the hour of the day with the highest average solar radiation based on the dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0035/835/35835284.ipynb_qa_2"
+kaggle_dataset_name = "dronio/SolarEnergy"
+gold_answer = "12"
+reward_mode_initial = "numeric"
+package_tier = 2
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "dronio__SolarEnergy"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "dronio/SolarEnergy"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "12"
+QUESTION = "What is the hour of the day with the highest average solar radiation based on the dataset?"
+REWARD_MODE = "numeric"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0036_230_36230590_qa_1/instruction.md

@@ -0,0 +1,15 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- (see /home/user/input)
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the third most used female name in the year 2002?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0036_230_36230590_qa_1/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0036_230_36230590_qa_1"
+description = "What is the third most used female name in the year 2002?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0036/230/36230590.ipynb_qa_1"
+kaggle_dataset_name = "kaggle/us-baby-names"
+gold_answer = "Hannah"
+reward_mode_initial = "exact_short"
+package_tier = 0
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "kaggle__us-baby-names"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "kaggle/us-baby-names"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "Hannah"
+QUESTION = "What is the third most used female name in the year 2002?"
+REWARD_MODE = "exact_short"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]

tasks/0037_236_37236169_qa_4/instruction.md

@@ -0,0 +1,19 @@
+You are a data-analysis agent working in a sandbox. Use your code-execution tool to inspect the files and compute the answer.
+
+Files (in /home/user/input, no subfolders):
+- PercentagePeopleBelowPovertyLevel.csv
+- PercentOver25CompletedHighSchool.csv
+- MedianHouseholdIncome2015.csv
+- PoliceKillingsUS.csv
+- ShareRaceByCity.csv
+
+Installed: pandas, numpy, matplotlib, seaborn, scipy, scikit-learn, statsmodels, tabulate, sqlite3, plotly (pip install more if needed).
+
+Question:
+What is the total number of unique names and surnames combined in the police killings dataset?
+
+Work it out step by step — inspect the data first (head, shape, dtypes), then compute.
+
+Answer with a single clean value: a bare number (no commas or units, e.g. 95293), a short label, yes/no, or a comma-separated list. Keep decimal precision. If there's no applicable answer, write: Not Applicable
+
+Write only that value to /workdir/answer.txt (e.g. `echo -n "<value>" > /workdir/answer.txt`), then stop.

tasks/0037_236_37236169_qa_4/task.toml

@@ -0,0 +1,63 @@
+schema_version = "1.2"
+artifacts = []
+
+[task]
+name = "data-agent-eval-v1/0037_236_37236169_qa_4"
+description = "What is the total number of unique names and surnames combined in the police killings dataset?"
+authors = []
+keywords = ["data-agent", "data-analysis", "kaggle"]
+
+[metadata]
+source_dataset = "jupyter-agent/jupyter-agent-dataset"
+source_row_id = "0037/236/37236169.ipynb_qa_4"
+kaggle_dataset_name = "kwullum/fatal-police-shootings-in-the-us"
+gold_answer = "4972"
+reward_mode_initial = "numeric"
+package_tier = 1
+difficulty_level = 0
+
+[environment]
+build_timeout_sec = 600.0
+os = "linux"
+cpus = 1
+memory_mb = 1024
+storage_mb = 5120
+gpus = 0
+allow_internet = true
+mcp_servers = []
+
+# Pre-agent hook: Harbor runs the command AFTER container start and BEFORE the
+# agent setup begins. We use it to pull this task's bucket prefix into
+# /home/user/input/. See environment/pull_bucket.py.
+[environment.healthcheck]
+command = "python3 /opt/pull_bucket.py && [ -n \"$(ls /home/user/input)\" ]"
+interval_sec = 2.0
+timeout_sec = 180.0
+start_period_sec = 5.0
+start_interval_sec = 2.0
+retries = 30
+
+[environment.env]
+HF_BUCKET = "AdithyaSK/jupyter-agent-kaggle-all"
+BUCKET_PREFIX = "kwullum__fatal-police-shootings-in-the-us"
+HF_TOKEN = "${HF_TOKEN}"
+KAGGLE_DATASET_NAME = "kwullum/fatal-police-shootings-in-the-us"
+
+[verifier]
+timeout_sec = 120.0
+
+[verifier.env]
+EXPECTED_ANSWER = "2646"
+QUESTION = "What is the total number of unique names and surnames combined in the police killings dataset?"
+REWARD_MODE = "numeric"
+ATOL = "0.0"
+RTOL = "0.0"
+
+[agent]
+# Capped at 600s (10 min) to kill the long-tail stuck-agent cases without
+# cutting off legitimate complex trials. Median Phase B trial is 60-120s;
+# legitimate L4/L5 tasks can hit 200-300s; anything past 600s is almost
+# certainly a stuck agent loop.
+timeout_sec = 600.0
+
+[solution.env]