Datasets:
messages stringlengths 156 36.2k | domain stringclasses 6
values | source stringclasses 5
values | config stringclasses 10
values | teacher stringclasses 8
values | verifier_passed bool 2
classes | ground_truth_json stringlengths 0 185 | n_tool_calls int32 0 17 | trace_kind stringclasses 4
values | disposition stringclasses 3
values | quality_score float32 0 1 | session_id stringlengths 0 64 | trajectory_id stringlengths 0 36 | category stringclasses 10
values | subcategory stringclasses 24
values | task stringlengths 0 100 | hermes_tool_trajectory bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[{"role":"system","content":"You are a grounded research assistant. Answer ONLY from the corpus and cite every claim in the requested format."},{"role":"user","content":"Using the corpus below, answer these questions. For each claim, cite sources inline as (doc-0001) style.\n1. What is the revenue of Cascade?\n2. How m... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | ["3840", "2207", "Nimbus", "60", "valves"] | 0 | single_turn | traj_success | 0 | false | |||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace and local corpus described by the user are synthetic ephemeral fixtures, not the user's live files or live services. Solve the task by using the provided tools through native OpenAI function calling (the assistant message mu... | tool | saidutta69/qwen-glm-kimi-distillation-clean | sft_tools | glm-5.2 | false | 7 | multi_turn_tool | 0 | false | |||||||
[{"role":"system","content":"You are a precise mathematician. Show brief steps, then end with exactly 'Answer: <value>'."},{"role":"user","content":"Compute 3/1 + 2/3 and simplify. End with: Answer: <value>"},{"role":"assistant","content":"3/1 + 2/3 = 9/3 + 2/3 = 11/3\n\nAnswer: 11/3"}] | math | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | "11/3" | 0 | single_turn | traj_success | 0 | false | |||||
[{"role":"system","content":"You are a precise structured-output generator. Output only the requested format, with no extra text, no code fences."},{"role":"user","content":"Create a YAML document that satisfies this schema. Invent realistic values.\nSchema: {\"additionalProperties\":false,\"properties\":{\"category\":... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | 0 | single_turn | traj_success | 0 | false | ||||||
[{"role":"system","content":"You are a personal assistant with a perfect memory for details given in this conversation. Always state the requested values exactly."},{"role":"user","content":"Monthly budget: utilities=$942, travel=$206, entertainment=$305."},{"role":"assistant","content":"Your monthly budget totals **$1... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | 0 | multi_turn | traj_success | 0 | false | ||||||
[{"role":"user","content":"<|im_start|>user\nWe intercepted a compiled Python function check_key(x) whose bytecode disassembly shows:\n 0 LOAD_FAST 0 (x)\n 2 LOAD_CONST 1 (68)\n 4 BINARY_OP 12 (^)\n 6 LOAD_CONST 2 (156)\n 8 COMPARE_OP 2 (==)\n ... | coding | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace and local corpus described by the user are synthetic ephemeral fixtures, not the user's live files or live services. Solve the task by using the provided tools through native OpenAI function calling (the assistant message mu... | tool | saidutta69/qwen-glm-kimi-distillation-clean | sft_tools | glm-5.2 | false | 4 | multi_turn_tool | 0 | false | |||||||
[{"role":"user","content":"Write a recipe for making lemon pepper chicken, a classic American dish. Provide details about the ingredients, cooking instructions, serving suggestions, and interesting facts related to the origins and popularity of lemon pepper chicken."},{"role":"assistant","content":"Ingredients:\n\n•4 c... | coding | saidutta69/fable-5-premium-v2 | agent_traces | claude-fable-5 | false | 0 | 1 | manus_2348575 | false | |||||||
[{"role":"user","content":"<start_of_turn>user\nPerform automated screenshot verification on rendered frame buffer #0113: verify that sky region luminance (203) exceeds terrain luminance (73), confirming correct atmospheric horizon rendering and zero black-screen shader crashes.<end_of_turn>\n<start_of_turn>model"},{"r... | agent | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"system","content":"You are an orchestration assistant. Follow the decomposition format exactly."},{"role":"user","content":"Produce a report on remote work tools.\nStep 1: decompose the work into exactly 3 parallel sub-tasks, each with a clear output description.\nStep 2: for each sub-task, write its output s... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | 0 | single_turn | traj_success | 0 | false | ||||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace and local corpus described by the user are synthetic ephemeral fixtures, not the user's live files or live services. Solve the task by using the provided tools through native OpenAI function calling (the assistant message mu... | tool | saidutta69/qwen-glm-kimi-distillation-clean | sft_tools | glm-5.2 | false | 3 | multi_turn_tool | 0 | false | |||||||
[{"role":"system","content":"Solve the synthetic mathematical task exactly. Use the stated parameters, show a check, and do not rely on outside benchmark examples.","tool_calls":[],"tool_call_id":"","name":""},{"role":"user","content":"Synthetic task seed (identity only): 10004447.\n\nDifferentiate f(x) = 6x^3. Return ... | math | saidutta69/qwen-glm-kimi-distillation-clean | sft_math | kimi-code/k3 | false | 0 | single_turn | 0 | false | |||||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace and local corpus described by the user are synthetic ephemeral fixtures, not the user's live files or live services. Solve the task by using the provided tools through native OpenAI function calling (the assistant message mu... | tool | saidutta69/qwen-glm-kimi-distillation-clean | sft_tools | glm-5.2 | false | 5 | multi_turn_tool | 0 | false | |||||||
[{"role":"user","content":"<|im_start|>user\nVerify that our WGSL Fractal Brownian Motion (FBM) heightmap compute function with 7 octaves and base frequency 0.015 stays strictly bounded within [-1.0, 1.0] at world coordinate (1687, 203) and produces no NaN values.<|im_end|>\n<|im_start|>assistant"},{"role":"assistant",... | coding | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"system","content":"You are a precise mathematician. Show brief steps, then end with exactly 'Answer: <value>'."},{"role":"user","content":"Evaluate: 25 + 12 × 10 - 5. Show your steps and end with: Answer: <value>"},{"role":"assistant","content":"25 + 12 × 10 - 5 \n= 25 + 120 - 5 \n= 140 \n\nAnswer: 140"}] | math | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | "140" | 0 | single_turn | traj_success | 0 | false | |||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace described by the user is a synthetic ephemeral fixture, not the user's live files. Solve the task using the provided tools through native function calling. Call tools with JSON arguments."},{"role":"user","content":"The work... | tool | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | 6 | multi_turn_tool | traj_success | 0 | false | ||||||
[{"role":"user","content":"<|im_start|>user\nWe have a raw EVM hex bytecode stream `600561046f55`. Write and execute a Python script to decode the opcodes (`PUSH1`, `PUSH2`, `SSTORE`) and inspect which storage slot is being modified.<|im_end|>\n<|im_start|>assistant"},{"role":"assistant","content":"<think>\nI will writ... | agent | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"user","content":"<start_of_turn>user\nWe have the following x86_64 stack frame setup:\n push rbp\n mov rbp, rsp\n sub rsp, 64\n mov QWORD PTR [rbp-40], 1483\n mov rax, QWORD PTR [rbp-40]\nWrite and execute a Python script to simulate this stack frame offset and verify the ret... | coding | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace described by the user is a synthetic ephemeral fixture, not the user's live files. Solve the task using the provided tools through native function calling. Call tools with JSON arguments."},{"role":"user","content":"The work... | tool | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | {"top_item": "calmvalves", "total": 995} | 7 | multi_turn_tool | traj_recovery | 0 | false | |||||
[{"role":"system","content":"You are a function calling AI model. You are provided with function signatures within <tools> </tools> XML tags. You may call one or more functions to assist with the user query. If available tools are not relevant in assisting with user query, just respond in natural conversational languag... | coding | lambda/hermes-agent-reasoning-traces | kimi | Kimi-K2.5 | false | 0 | 0 | 7e574ea1-bb56-4426-a626-36be1d2e15fa | Terminal & Coding | Terminal Tasks | Set up GitHub Actions CI for a VS Code extension project with linting and testing | true | ||||
[{"role":"user","content":"<|im_start|>user\nOur distributed worker pool (`pool_229.py`) suffers from a subtle race condition under concurrency load. When running 13 async workers incrementing a shared state counter 729 times each, the final total frequently falls below the expected invariant (9477). Write an executabl... | coding | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"user","content":"<|im_start|>user\nWe have a raw EVM hex bytecode stream `600861040455`. Write and execute a Python script to decode the opcodes (`PUSH1`, `PUSH2`, `SSTORE`) and inspect which storage slot is being modified.<|im_end|>\n<|im_start|>assistant"},{"role":"assistant","content":"<think>\nI will writ... | agent | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"system","content":"You are a careful deterministic workspace assistant. The workspace described by the user is a synthetic ephemeral fixture, not the user's live files. Solve the task using the provided tools through native function calling. Call tools with JSON arguments."},{"role":"user","content":"Perform ... | tool | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | 6 | multi_turn_tool | traj_success | 0 | false | ||||||
[{"role":"user","content":"Discuss how Clubs in day-to-day life are defined and what members they consist of, including an example of a specific club and an analysis of how that club's members and rules satisfy the requirements to be called a club."},{"role":"assistant","content":"Members: Clubs consist of members who ... | agent | saidutta69/fable-5-premium-v2 | agent_traces | claude-fable-5 | false | 0 | 1 | manus_2379092 | false | |||||||
[{"role":"system","content":"You are a precise structured-output generator. Output only the requested format, with no extra text, no code fences."},{"role":"user","content":"Parse the following document and populate the YAML data structure per the schema. Take values ONLY from the document.\nSchema: {\"additionalProper... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | {"city": "Novaatlas", "length": 271, "level": "critical", "name": "Kitemaple", "requiresReview": true, "size": 431} | 0 | single_turn | traj_success | 0 | false | |||||
[{"role":"system","content":"You are a planning assistant. ALWAYS output a numbered plan first, then the complete deliverable."},{"role":"user","content":"Design a data pipeline that ingests sales data, cleans it, aggregates it, and exports a report.\n\nFirst, write a numbered plan with 3-5 steps. Then, in the same res... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | ["ingest", "clean", "aggregate", "export"] | 0 | single_turn | traj_success | 0 | false | |||||
[{"role":"system","content":"You are an expert software engineer. Write complete, correct, runnable code in the language requested by the user. Follow the requested language, interface, and constraints. Explain the approach briefly when requested.","tool_calls":[],"tool_call_id":"","name":""},{"role":"user","content":"... | coding | saidutta69/qwen-glm-kimi-distillation-clean | sft_code | qwen3.8-max-preview | false | 0 | single_turn | 0 | false | |||||||
[{"role":"user","content":"Explain an approach for identifying and fixing logical errors in JavaScript code."},{"role":"assistant","content":"1. Understand the intended behavior. Clarify the expected output and behavior of the code by discussing it with the author or checking documentation. This will establish a baseli... | coding | saidutta69/fable-5-premium-v2 | agent_traces | claude-fable-5 | false | 0 | 1 | manus_2328527 | false | |||||||
[{"role":"user","content":"Are there any animals that look very similar but are not closely related?"},{"role":"assistant","content":"In evolutionary biology, the phenomenon where unrelated or distantly related organisms evolve similar physical traits is called **Convergent Evolution**. \n\nThis occurs when different s... | coding | saidutta69/fable-5-premium-v2 | agent_traces | claude-fable-5 | false | 0 | 1 | manus_2381450 | false | |||||||
[{"role":"system","content":"You are a function calling AI model. You are provided with function signatures within <tools> </tools> XML tags. You may call one or more functions to assist with the user query. If available tools are not relevant in assisting with user query, just respond in natural conversational languag... | coding | lambda/hermes-agent-reasoning-traces | kimi | Kimi-K2.5 | false | 0 | 0 | 69b846b1-d327-4ed6-ad28-b1802c25f25e | Terminal & Coding | Terminal Tasks | Debug this error I'm getting when running `docker-compose up`: FileNotFoundError: config | true | ||||
[{"role":"user","content":"<|im_start|>user\nPerform automated screenshot verification on rendered frame buffer #0251: verify that sky region luminance (191) exceeds terrain luminance (61), confirming correct atmospheric horizon rendering and zero black-screen shader crashes.<|im_end|>\n<|im_start|>assistant"},{"role":... | agent | True2456/code-as-action | train | frontier-generated | false | 0 | 0 | false | ||||||||
[{"role":"system","content":"You are an expert software engineer. Write complete, correct, runnable code in the language requested by the user. Follow the requested language, interface, and constraints. Explain the approach briefly when requested.","tool_calls":[],"tool_call_id":"","name":""},{"role":"user","content":"... | coding | saidutta69/qwen-glm-kimi-distillation-clean | sft_code | qwen3.8-max-preview | false | 0 | single_turn | 0 | false | |||||||
[{"role":"system","content":"You are a personal assistant with a perfect memory for details given in this conversation. Always state the requested values exactly."},{"role":"user","content":"I have meetings on Tuesday (12:00), Wednesday (13:00), and Thursday (13:00)."},{"role":"assistant","content":"Noted. You have mee... | agent | r0b0tlab/deepseek-v4-pro-0813-agentic | canonical | deepseek-v4-pro | true | 0 | multi_turn | traj_success | 0 | false | ||||||
[{"role":"system","content":"You are an expert mathematician. Solve the given problem with complete rigor.\n\nFormat your response EXACTLY as:\n<think>\n[Your complete step-by-step reasoning. Show all intermediate calculations.\nIf you make an error and catch it, show the correction.\nEnd with a clear statement of the ... | math | saidutta69/qwen-glm-kimi-distillation-clean | sft_math | qwen3.8-max-preview | false | 0 | single_turn | 0 | false | |||||||
[{"role":"system","content":"You are a function calling AI model. You are provided with function signatures within <tools> </tools> XML tags. You may call one or more functions to assist with the user query. If available tools are not relevant in assisting with user query, just respond in natural conversational languag... | coding | lambda/hermes-agent-reasoning-traces | glm-5.1 | GLM-5.1 | false | 0 | 0 | e45ca833-8598-49b7-8cd0-72151c7d8345 | Terminal & Coding | Terminal Tasks | Use the OCR and documents skill to extract text from a screenshot of an error message. | true |
End of preview. Expand in Data Studio
Agentic Dataset v6 Dynamic
Dynamically allocated SFT dataset.
Requested target: 100,000
Actual target: 100,000
Domain distribution
coding: 32,200 agent: 15,235 tool: 22,565 math: 14,000 instruction: 10,000 stem: 6,000
Reliability
anti_hallucination: 30,512
tool_call_rows: 22,265
tool_result_rows: 21,972
tool_grounded_rows: 21,971
verified_rows: 18,384
uncertainty_rows: 1,058
fake_tool_claims: 0
Length
P50 chars: 2,038
P95 chars: 14,241
P99 chars: 26,783
The anti-hallucination flag is a training-data quality signal, not a guarantee of zero hallucinations at inference time.
- Downloads last month
- 81