Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
51
85
record_type
stringclasses
1 value
grade_level
stringclasses
20 values
subject
stringclasses
10 values
concept_id
stringlengths
27
53
title
stringlengths
27
53
text
stringlengths
396
2.99k
messages
stringlengths
1.36k
4.26k
prompt
stringlengths
396
2.99k
response
stringlengths
627
722
chosen
stringlengths
627
722
rejected
stringclasses
1 value
rubric
stringclasses
10 values
scoring_dimensions
listlengths
4
11
metadata
stringlengths
2.89k
6.32k
source_file
stringclasses
1 value
PK_MATH_SORTING_PATTERNS_AND_NUMBER_L4_T1_TOOL_ROUTER_TRACE
agent_tool_traces
Pre-K
Mathematics
PK_MATH_SORTING_PATTERNS_AND_NUMBER
Pk Math Sorting Patterns And Number
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Sorting, Patterns, and Number Sense in Mathematics, decide whether to answer directly or use tools. Candidate tools: calculator.evaluate, rubric.score. Return a plan ...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "The user needs a reliable answer, but unnecessary tool use should be avoi...
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Sorting, Patterns, and Number Sense in Mathematics, decide whether to answer directly or use tools. Candidate tools: calculator.evaluate, rubric.score. Return a plan ...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["calculator.evaluate", "rubric.score"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answer for Sorting, Patterns, and N...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["calculator.evaluate", "rubric.score"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answer for Sorting, Patterns, and N...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["calculator.evaluate", "rubric.score"], "concept_id": "PK_MATH_SORTING_PATTERNS_AND_NUMBER", "grade_level": "Pre-K", "raw_row": {"id": "PK_MATH_SORTING_PATTERNS_AND_NUMBER_L4_T1_TOOL_ROUTER_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools only when they reduce uncertaint...
agent_tool_traces.jsonl
PK_MATH_SORTING_PATTERNS_AND_NUMBER_L4_T2_FUNCTION_CALL_TRACE_TRACE
agent_tool_traces
Pre-K
Mathematics
PK_MATH_SORTING_PATTERNS_AND_NUMBER
Pk Math Sorting Patterns And Number
{"available_tools": [{"tool_name": "calculator.evaluate", "category": "computation", "description": "Evaluate arithmetic, algebraic, statistical, or unit-conversion expressions.", "input_schema": {"expression": "string", "units": "optional string", "precision": "optional integer"}, "output_schema": {"result": "number o...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "{\"available_tools\": [{\"tool_name\": \"calculator.evaluate\", \"categor...
{"available_tools": [{"tool_name": "calculator.evaluate", "category": "computation", "description": "Evaluate arithmetic, algebraic, statistical, or unit-conversion expressions.", "input_schema": {"expression": "string", "units": "optional string", "precision": "optional integer"}, "output_schema": {"result": "number o...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["calculator.evaluate", "rubric.score"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answer for Sorting, Patterns, and ...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["calculator.evaluate", "rubric.score"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answer for Sorting, Patterns, and ...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "self_correction", "self_correction_behavior", "emotional_understanding" ]
{"available_tools": ["calculator.evaluate", "rubric.score"], "concept_id": "PK_MATH_SORTING_PATTERNS_AND_NUMBER", "grade_level": "Pre-K", "raw_row": {"id": "PK_MATH_SORTING_PATTERNS_AND_NUMBER_L4_T2_FUNCTION_CALL_TRACE_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools only when they reduce un...
agent_tool_traces.jsonl
PK_MATH_SORTING_PATTERNS_AND_NUMBER_L4_T3_TOOL_RESULT_AUDIT_TRACE
agent_tool_traces
Pre-K
Mathematics
PK_MATH_SORTING_PATTERNS_AND_NUMBER
Pk Math Sorting Patterns And Number
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Sorting, Patterns, and Number Sense. Identify whether the result is sufficient, suspicious, incomplet...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "Hypothetical observation: a tool returned a confident result but did not ...
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Sorting, Patterns, and Number Sense. Identify whether the result is sufficient, suspicious, incomplet...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["calculator.evaluate", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_answer": "A high-quality answe...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["calculator.evaluate", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_answer": "A high-quality answe...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["calculator.evaluate", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_MATH_SORTING_PATTERNS_AND_NUMBER", "grade_level": "Pre-K", "raw_row": {"id": "PK_MATH_SORTING_PATTERNS_AND_NUMBER_L4_T3_TOOL_RESULT_AUDIT_TRACE", "messages": [{"content": "You are an industry-grade agent. Use...
agent_tool_traces.jsonl
PK_ELA_ORAL_LANGUAGE_STORY_LISTENING_L4_T1_TOOL_ROUTER_TRACE
agent_tool_traces
Pre-K
English Language Arts
PK_ELA_ORAL_LANGUAGE_STORY_LISTENING
Pk Ela Oral Language Story Listening
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Oral Language, Story Listening, and Print Awareness in English Language Arts, decide whether to answer directly or use tools. Candidate tools: retrieval.search, rubri...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "The user needs a reliable answer, but unnecessary tool use should be avoi...
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Oral Language, Story Listening, and Print Awareness in English Language Arts, decide whether to answer directly or use tools. Candidate tools: retrieval.search, rubri...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answe...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answe...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_ELA_ORAL_LANGUAGE_STORY_LISTENING", "grade_level": "Pre-K", "raw_row": {"id": "PK_ELA_ORAL_LANGUAGE_STORY_LISTENING_L4_T1_TOOL_ROUTER_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools ...
agent_tool_traces.jsonl
PK_ELA_ORAL_LANGUAGE_STORY_LISTENING_L4_T2_FUNCTION_CALL_TRACE_TRACE
agent_tool_traces
Pre-K
English Language Arts
PK_ELA_ORAL_LANGUAGE_STORY_LISTENING
Pk Ela Oral Language Story Listening
{"available_tools": [{"tool_name": "retrieval.search", "category": "retrieval", "description": "Search trusted sources or a local corpus for evidence.", "input_schema": {"query": "string", "source_policy": "trusted_only | local_only | broad_web", "recency_required": "boolean"}, "output_schema": {"sources": "list[object...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "{\"available_tools\": [{\"tool_name\": \"retrieval.search\", \"category\"...
{"available_tools": [{"tool_name": "retrieval.search", "category": "retrieval", "description": "Search trusted sources or a local corpus for evidence.", "input_schema": {"query": "string", "source_policy": "trusted_only | local_only | broad_web", "recency_required": "boolean"}, "output_schema": {"sources": "list[object...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answ...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answ...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "self_correction", "judgment", "self_correction_behavior", "emotional_understanding", "stewardship", "value_balancing" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_ELA_ORAL_LANGUAGE_STORY_LISTENING", "grade_level": "Pre-K", "raw_row": {"id": "PK_ELA_ORAL_LANGUAGE_STORY_LISTENING_L4_T2_FUNCTION_CALL_TRACE_TRACE", "messages": [{"content": "You are an industry-grade agent. Us...
agent_tool_traces.jsonl
PK_ELA_ORAL_LANGUAGE_STORY_LISTENING_L4_T3_TOOL_RESULT_AUDIT_TRACE
agent_tool_traces
Pre-K
English Language Arts
PK_ELA_ORAL_LANGUAGE_STORY_LISTENING
Pk Ela Oral Language Story Listening
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Oral Language, Story Listening, and Print Awareness. Identify whether the result is sufficient, suspi...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "Hypothetical observation: a tool returned a confident result but did not ...
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Oral Language, Story Listening, and Print Awareness. Identify whether the result is sufficient, suspi...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review", "human_escalate.request_review"], "concept_id": "PK_ELA_ORAL_LANGUAGE_STORY_LISTENING", "grade_level": "Pre-K", "raw_row": {"id": "PK_ELA_ORAL_LANGUAGE_STORY_LISTENING_L4_T3_TOOL_RESULT_AUDIT_TRACE", "messages": [{"content": "You ...
agent_tool_traces.jsonl
PK_SCI_OBSERVATION_OF_LIVING_AND_L4_T1_TOOL_ROUTER_TRACE
agent_tool_traces
Pre-K
Science
PK_SCI_OBSERVATION_OF_LIVING_AND
Pk Sci Observation Of Living And
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Observation of Living and Nonliving Things in Science, decide whether to answer directly or use tools. Candidate tools: retrieval.search, calculator.evaluate, citatio...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "The user needs a reliable answer, but unnecessary tool use should be avoi...
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Observation of Living and Nonliving Things in Science, decide whether to answer directly or use tools. Candidate tools: retrieval.search, calculator.evaluate, citatio...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answer for Ob...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answer for Ob...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit"], "concept_id": "PK_SCI_OBSERVATION_OF_LIVING_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_SCI_OBSERVATION_OF_LIVING_AND_L4_T1_TOOL_ROUTER_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools only when they r...
agent_tool_traces.jsonl
PK_SCI_OBSERVATION_OF_LIVING_AND_L4_T2_FUNCTION_CALL_TRACE_TRACE
agent_tool_traces
Pre-K
Science
PK_SCI_OBSERVATION_OF_LIVING_AND
Pk Sci Observation Of Living And
{"available_tools": [{"tool_name": "calculator.evaluate", "category": "computation", "description": "Evaluate arithmetic, algebraic, statistical, or unit-conversion expressions.", "input_schema": {"expression": "string", "units": "optional string", "precision": "optional integer"}, "output_schema": {"result": "number o...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "{\"available_tools\": [{\"tool_name\": \"calculator.evaluate\", \"categor...
{"available_tools": [{"tool_name": "calculator.evaluate", "category": "computation", "description": "Evaluate arithmetic, algebraic, statistical, or unit-conversion expressions.", "input_schema": {"expression": "string", "units": "optional string", "precision": "optional integer"}, "output_schema": {"result": "number o...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answer for O...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answer for O...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "judgment", "emotional_understanding", "stewardship", "value_balancing" ]
{"available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit"], "concept_id": "PK_SCI_OBSERVATION_OF_LIVING_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_SCI_OBSERVATION_OF_LIVING_AND_L4_T2_FUNCTION_CALL_TRACE_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools only whe...
agent_tool_traces.jsonl
PK_SCI_OBSERVATION_OF_LIVING_AND_L4_T3_TOOL_RESULT_AUDIT_TRACE
agent_tool_traces
Pre-K
Science
PK_SCI_OBSERVATION_OF_LIVING_AND
Pk Sci Observation Of Living And
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Observation of Living and Nonliving Things. Identify whether the result is sufficient, suspicious, in...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "Hypothetical observation: a tool returned a confident result but did not ...
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Observation of Living and Nonliving Things. Identify whether the result is sufficient, suspicious, in...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_answer":...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_answer":...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["retrieval.search", "calculator.evaluate", "citation.audit", "human_escalate.request_review"], "concept_id": "PK_SCI_OBSERVATION_OF_LIVING_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_SCI_OBSERVATION_OF_LIVING_AND_L4_T3_TOOL_RESULT_AUDIT_TRACE", "messages": [{"content": "You are an industry-...
agent_tool_traces.jsonl
PK_SOC_FAMILY_COMMUNITY_HELPERS_AND_L4_T1_TOOL_ROUTER_TRACE
agent_tool_traces
Pre-K
Social Studies and Civics
PK_SOC_FAMILY_COMMUNITY_HELPERS_AND
Pk Soc Family Community Helpers And
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Family, Community Helpers, and Fairness in Social Studies and Civics, decide whether to answer directly or use tools. Candidate tools: retrieval.search, rubric.score,...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "The user needs a reliable answer, but unnecessary tool use should be avoi...
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Family, Community Helpers, and Fairness in Social Studies and Civics, decide whether to answer directly or use tools. Candidate tools: retrieval.search, rubric.score,...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answe...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answe...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "judgment", "perspective_taking", "emotional_understanding", "value_balancing" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_SOC_FAMILY_COMMUNITY_HELPERS_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_SOC_FAMILY_COMMUNITY_HELPERS_AND_L4_T1_TOOL_ROUTER_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools on...
agent_tool_traces.jsonl
PK_SOC_FAMILY_COMMUNITY_HELPERS_AND_L4_T2_FUNCTION_CALL_TRACE_TRACE
agent_tool_traces
Pre-K
Social Studies and Civics
PK_SOC_FAMILY_COMMUNITY_HELPERS_AND
Pk Soc Family Community Helpers And
{"available_tools": [{"tool_name": "retrieval.search", "category": "retrieval", "description": "Search trusted sources or a local corpus for evidence.", "input_schema": {"query": "string", "source_policy": "trusted_only | local_only | broad_web", "recency_required": "boolean"}, "output_schema": {"sources": "list[object...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "{\"available_tools\": [{\"tool_name\": \"retrieval.search\", \"category\"...
{"available_tools": [{"tool_name": "retrieval.search", "category": "retrieval", "description": "Search trusted sources or a local corpus for evidence.", "input_schema": {"query": "string", "source_policy": "trusted_only | local_only | broad_web", "recency_required": "boolean"}, "output_schema": {"sources": "list[object...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answ...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answ...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "self_correction", "judgment", "perspective_taking", "self_correction_behavior", "emotional_understanding", "stewardship", "value_balancing" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_SOC_FAMILY_COMMUNITY_HELPERS_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_SOC_FAMILY_COMMUNITY_HELPERS_AND_L4_T2_FUNCTION_CALL_TRACE_TRACE", "messages": [{"content": "You are an industry-grade agent. Use ...
agent_tool_traces.jsonl
PK_SOC_FAMILY_COMMUNITY_HELPERS_AND_L4_T3_TOOL_RESULT_AUDIT_TRACE
agent_tool_traces
Pre-K
Social Studies and Civics
PK_SOC_FAMILY_COMMUNITY_HELPERS_AND
Pk Soc Family Community Helpers And
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Family, Community Helpers, and Fairness. Identify whether the result is sufficient, suspicious, incom...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "Hypothetical observation: a tool returned a confident result but did not ...
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Family, Community Helpers, and Fairness. Identify whether the result is sufficient, suspicious, incom...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "judgment", "perspective_taking", "emotional_understanding", "value_balancing" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review", "human_escalate.request_review"], "concept_id": "PK_SOC_FAMILY_COMMUNITY_HELPERS_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_SOC_FAMILY_COMMUNITY_HELPERS_AND_L4_T3_TOOL_RESULT_AUDIT_TRACE", "messages": [{"content": "You ar...
agent_tool_traces.jsonl
PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL_L4_T1_TOOL_ROUTER_TRACE
agent_tool_traces
Pre-K
Computer Science and AI
PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL
Pk Csai Sequencing Algorithms And Helpful
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Sequencing, Algorithms, and Helpful Machines in Computer Science and AI, decide whether to answer directly or use tools. Candidate tools: code.run_tests, dataset.quer...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "The user needs a reliable answer, but unnecessary tool use should be avoi...
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Sequencing, Algorithms, and Helpful Machines in Computer Science and AI, decide whether to answer directly or use tools. Candidate tools: code.run_tests, dataset.quer...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["code.run_tests", "dataset.query", "rubric.score"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answer for Sequencing, ...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["code.run_tests", "dataset.query", "rubric.score"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answer for Sequencing, ...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["code.run_tests", "dataset.query", "rubric.score"], "concept_id": "PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL", "grade_level": "Pre-K", "raw_row": {"id": "PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL_L4_T1_TOOL_ROUTER_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools only whe...
agent_tool_traces.jsonl
PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL_L4_T2_FUNCTION_CALL_TRACE_TRACE
agent_tool_traces
Pre-K
Computer Science and AI
PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL
Pk Csai Sequencing Algorithms And Helpful
{"available_tools": [{"tool_name": "code.run_tests", "category": "execution", "description": "Run code, tests, static checks, or small simulations in a controlled environment.", "input_schema": {"language": "string", "code_or_command": "string", "test_goal": "string", "timeout_seconds": "integer"}, "output_schema": {"s...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "{\"available_tools\": [{\"tool_name\": \"code.run_tests\", \"category\": ...
{"available_tools": [{"tool_name": "code.run_tests", "category": "execution", "description": "Run code, tests, static checks, or small simulations in a controlled environment.", "input_schema": {"language": "string", "code_or_command": "string", "test_goal": "string", "timeout_seconds": "integer"}, "output_schema": {"s...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["code.run_tests", "dataset.query", "rubric.score"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answer for Sequencing,...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["code.run_tests", "dataset.query", "rubric.score"], "expected_trace_shape": ["thought_summary", "tool_call", "observation", "verification", "final"], "final_answer": "A high-quality answer for Sequencing,...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "self_correction", "self_correction_behavior", "emotional_understanding" ]
{"available_tools": ["code.run_tests", "dataset.query", "rubric.score"], "concept_id": "PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL", "grade_level": "Pre-K", "raw_row": {"id": "PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL_L4_T2_FUNCTION_CALL_TRACE_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools ...
agent_tool_traces.jsonl
PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL_L4_T3_TOOL_RESULT_AUDIT_TRACE
agent_tool_traces
Pre-K
Computer Science and AI
PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL
Pk Csai Sequencing Algorithms And Helpful
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Sequencing, Algorithms, and Helpful Machines. Identify whether the result is sufficient, suspicious, ...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "Hypothetical observation: a tool returned a confident result but did not ...
Hypothetical observation: a tool returned a confident result but did not show source date, method, units, or sample quality. Treat this as a realistic industry reliability problem. Audit a hypothetical tool result for Sequencing, Algorithms, and Helpful Machines. Identify whether the result is sufficient, suspicious, ...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["code.run_tests", "dataset.query", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_answer": "A high-q...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["code.run_tests", "dataset.query", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["inspect_observation", "risk_check", "verify_or_escalate", "final"], "final_answer": "A high-q...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["code.run_tests", "dataset.query", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL", "grade_level": "Pre-K", "raw_row": {"id": "PK_CSAI_SEQUENCING_ALGORITHMS_AND_HELPFUL_L4_T3_TOOL_RESULT_AUDIT_TRACE", "messages": [{"content": "You are an i...
agent_tool_traces.jsonl
PK_ART_COLOR_SOUND_MOVEMENT_AND_L4_T1_TOOL_ROUTER_TRACE
agent_tool_traces
Pre-K
Arts and Music
PK_ART_COLOR_SOUND_MOVEMENT_AND
Pk Art Color Sound Movement And
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Color, Sound, Movement, and Creative Choice in Arts and Music, decide whether to answer directly or use tools. Candidate tools: retrieval.search, rubric.score, human_...
[{"content": "You are an industry-grade agent. Use tools only when they reduce uncertainty or verify work. Provide concise planning summaries, valid tool arguments, observations, verification, and a final answer.", "role": "system"}, {"content": "The user needs a reliable answer, but unnecessary tool use should be avoi...
The user needs a reliable answer, but unnecessary tool use should be avoided. Assume tool calls have cost, latency, and possible error. For a task about Color, Sound, Movement, and Creative Choice in Arts and Music, decide whether to answer directly or use tools. Candidate tools: retrieval.search, rubric.score, human_...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answe...
{"plan_summary": "Define goal, choose minimum useful tools, verify observations, then answer.", "available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "expected_trace_shape": ["reason", "select_tool", "call_or_skip", "observe", "verify", "final"], "final_answer": "A high-quality answe...
Not applicable for this record type.
{"accuracy": "0.0-1.0 for correctness, factual fit, and completion of required work.", "emotional_understanding": "0.0-1.0 for recognizing emotions and responding proportionally.", "explanation_quality": "0.0-1.0 for clear reasoning, evidence, and useful structure.", "judgment": "0.0-1.0 for defensible decisions under ...
[ "accuracy", "explanation_quality", "uncertainty_awareness", "emotional_understanding" ]
{"available_tools": ["retrieval.search", "rubric.score", "human_escalate.request_review"], "concept_id": "PK_ART_COLOR_SOUND_MOVEMENT_AND", "grade_level": "Pre-K", "raw_row": {"id": "PK_ART_COLOR_SOUND_MOVEMENT_AND_L4_T1_TOOL_ROUTER_TRACE", "messages": [{"content": "You are an industry-grade agent. Use tools only when ...
agent_tool_traces.jsonl
End of preview. Expand in Data Studio
FOUNDATION DATASET AI Mastery Foundation Curriculum A premium foundation layer for knowledge, reasoning, preference, reward, benchmark, and agentic tool-use training.
Hugging Face-ready Parquet package

AI Mastery Foundation Curriculum

A premium staged foundation dataset for building models with a cleaner first layer of academic knowledge, reasoning discipline, preference judgment, benchmark awareness, and tool-use behavior before they move into larger downstream corpora.

31,331Total packaged rows
20Grade bands
10Subjects per grade
0Duplicate IDs
0Duplicate prompt-response pairs
8HF configs

A foundation layer built for serious training pipelines

Most datasets are raw piles of examples. This one is structured like a training system: concepts first, supervised answers next, preference contrasts after that, reward references for scoring, benchmark rows for evaluation, and agentic traces for tool-aware behavior. It is designed to make the earliest stage of a model pipeline feel less chaotic and more curriculum-shaped.

Curriculum-first Rubric-backed Preference-ready Benchmark-aware Agentic-tool-use ready Modern HF Parquet

What This Dataset Is

AI Mastery Foundation Curriculum is a staged training and evaluation package spanning Pre-K, K-12, undergraduate, graduate, and doctoral learning. It is organized for practical AI training workflows: curriculum grounding, supervised fine-tuning, preference optimization, reward-model/reference judging, benchmark evaluation, and agentic tool-use behavior. The goal is not just to provide rows, but to provide a progression: what the model should know, how it should answer, how better answers differ from weaker ones, and how performance should be checked.
Curriculum graph SFT conversations RLHF pairs Reward references Benchmarks Agentic traces Curated expansion
Design intent

The package is built as a foundation layer: broad enough to cover academic basics, structured enough for training pipelines, and clean enough to be mixed with later domain corpora. Each major training surface keeps prompts, responses, rubrics, scoring dimensions, titles, and metadata aligned so the dataset can support training, judging, retrieval, filtering, and inspection.

Recommended Use

Use this as an early foundation and alignment-oriented curriculum layer: first expose a model or training pipeline to clean educational structure, worked reasoning, expected answers, rubrics, and preference contrasts; then continue into larger domain-specific or general-purpose corpora. The dataset is also useful for model evaluation because the benchmark, reward, and agentic configs keep task format, rubric, and metadata consistent.
Before large-scale tuning

Use it as a structured primer so early training sees ordered knowledge instead of only loose internet-style text.

During alignment work

Use preference, reward, and rubric fields to connect answer quality with explicit scoring criteria.

During evaluation

Use held-out benchmark and agentic configs to track reasoning, uncertainty, transfer, and tool behavior.

During dataset ops

Use the shared schema, titles, metadata, and Parquet layout for filtering, dashboards, audits, and mixing.

Training Stage Map

Stage 01Curriculum Grounding

Use curriculum to expose the model or indexing layer to grade, subject, concept, prerequisite, misconception, and objective structure.

Stage 02Supervised Reasoning

Use sft for instruction-following, worked answers, mistake correction, and grade-appropriate response style.

Stage 03Preference Alignment

Use rlhf to teach contrastive judgment between strong and weak answers with explicit preference reasons.

Stage 04Reward and Judging

Use reward for rubric-backed references, scalar scoring setups, or LLM-judge calibration.

Stage 05Benchmark Evaluation

Use benchmark as held-out evaluation across mastery, transfer, reflection, and capstone-style tasks.

Stage 06Agentic Behavior

Use agentic for tool routing, trace shape, verification, uncertainty, and escalation behavior.

Config Purpose Guide

all

Unified table for broad mixing, inspection, sampling, and lightweight baseline training.

curriculum

Concept-level map across grade bands, subjects, objectives, prerequisites, and misconception metadata.

sft

Conversation-style supervised rows for instruction following and answer construction.

rlhf

Chosen and rejected answers for preference optimization and reward comparisons.

reward

Reference-answer rows with rubrics and scoring dimensions for judge or reward-model work.

benchmark

Held-out test split for capability tracking, regression checks, and curriculum coverage audits.

agentic

Tool-use traces and agentic test items for routing, verification, and escalation behavior.

curated_expansion

Large cleaned expansion of examples and solutions folded into the package as normal dataset rows.

Configs and Counts

ConfigSplit/FileRows
allall/train.parquet31,331
curriculumcurriculum/train.parquet200
sftsft/train.parquet2,730
sftsft/validation.parquet270
rlhfrlhf/train.parquet1,003
rewardreward/train.parquet3,000
benchmarkbenchmark/test.parquet1,240
agenticagentic/train.parquet600
agenticagentic/test.parquet600
curated_expansioncurated_expansion/train.parquet21,688

Schema

Every split uses the same schema so configs can be combined safely:

id, record_type, grade_level, subject, concept_id, title, text, messages, prompt, response, chosen, rejected, rubric, scoring_dimensions, metadata, source_file.

Nested details are preserved as JSON text in metadata where useful. The shared schema makes the package simple to load with the modern Hugging Face dataset-card config system.

id
Stable row identifier.
record_type
Config/source role.
grade_level
Pre-K through PhD.
subject
Academic domain.
prompt
Task or instruction.
response
Reference answer.
chosen
Preferred answer when available.
rejected
Contrastive weak answer.
rubric
JSON scoring guide.
metadata
JSON route details.

Loading

from datasets import load_dataset

sft = load_dataset("./hf_dataset", "sft")
curated = load_dataset("./hf_dataset", "curated_expansion")
agentic = load_dataset("./hf_dataset", "agentic")

Direct Parquet loading also works:

from datasets import load_dataset

ds = load_dataset("parquet", data_files={"train": "./hf_dataset/all/train.parquet"})

Quality Status

Verified loadingAll configs load with current Hugging Face datasets using load_dataset("./hf_dataset", config).
Clean IDsDuplicate IDs are 0 in every packaged split.
Clean pairsDuplicate prompt-response pairs are 0 in every packaged split.
Portable formatThe public package uses standard Parquet files and dataset-card YAML configs.

Workflow Fit

Foundation pass

Sample curriculum, sft, and curated_expansion to build broad educational coverage before heavier domain corpora.

Alignment pass

Mix rlhf and reward when training preference, scoring, ranking, or judge components.

Evaluation pass

Use benchmark and agentic/test as held-out checks for reasoning, transfer, tool use, uncertainty, and escalation.

Inspection pass

Use all for search, filtering, dashboards, metadata QA, and unified analytics over the entire package.

Creation Provenance

The dataset is authored, selected, structured, cleaned, and packaged as the AI Mastery Curriculum dataset. Some expansion drafting used the Gemini API during local development, then the resulting rows were curated, deduplicated, renumbered, normalized, and repackaged into this publishable dataset format.

Downloads last month
31