id large_stringlengths 5 8 | kind large_stringclasses 2
values | tags listlengths 1 4 | tools listlengths 0 4 | messages listlengths 3 7 | meta dict | text large_stringlengths 1.04k 8.2k |
|---|---|---|---|---|---|---|
S69-S2 | tool | [
"data",
"filter"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-filter",
"world": "NanMedia Manufacturing",
"industry": "retail",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S5-N4 | tool | [
"search"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-search",
"world": "SomCare Co. Ltd.",
"industry": "e-commerce",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S141-D0 | tool | [
"search",
"data",
"verify",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D9-cross-check",
"world": "OriGold Co. Ltd.",
"industry": "manufacturing",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S27-D4 | tool | [
"data",
"filter",
"stats",
"multi"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D4-filter-stats",
"world": "NanSummit Partners",
"industry": "logistics",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S157-D8 | tool | [
"data",
"filter",
"stats",
"multi"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D4-filter-stats",
"world": "FonLogix Partners",
"industry": "construction",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S127-D9 | tool | [
"data",
"aggregate",
"multi"
] | [
{
"type": "function",
"function": {
"name": "csv_head",
"description": "Show the first N rows of a CSV table.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": null,
"op": null,
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D3-head-aggregate",
"world": "ItoWorks Solutions",
"industry": "agribusiness",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S164-D9 | tool | [
"read",
"sections",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_sections",
"description": "List the section headings of a document (table of contents).",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": nul... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D6-sections-read",
"world": "HapUnion Trading",
"industry": "agribusiness",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S43-D13 | tool | [
"browse",
"metadata",
"multi"
] | [
{
"type": "function",
"function": {
"name": "folder_list",
"description": "List documents in a folder.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": null,
"op": null,
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D5-folder-metadata",
"world": "KaoGoods Co., Ltd.",
"industry": "professional services",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S101-D10 | tool | [
"data",
"aggregate",
"multi"
] | [
{
"type": "function",
"function": {
"name": "csv_head",
"description": "Show the first N rows of a CSV table.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": null,
"op": null,
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D3-head-aggregate",
"world": "NokMedia Trading",
"industry": "food & beverage",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S99-S7 | tool | [
"data",
"filter"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-filter",
"world": "BenPrime Co., Ltd.",
"industry": "construction",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S50-D5 | tool | [
"read",
"sections",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_sections",
"description": "List the section headings of a document (table of contents).",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": nul... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D6-sections-read",
"world": "ChaiWear Enterprises",
"industry": "retail",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S96-S2 | tool | [
"read"
] | [
{
"type": "function",
"function": {
"name": "doc_read",
"description": "Read the full text of one document by path.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-read",
"world": "OriGreen Manufacturing",
"industry": "real estate",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S21-D7 | tool | [
"read",
"sections",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_sections",
"description": "List the section headings of a document (table of contents).",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": nul... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D6-sections-read",
"world": "SanRidge Public Company",
"industry": "professional services",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S93-D11 | tool | [
"data",
"aggregate",
"compare",
"multi"
] | [
{
"type": "function",
"function": {
"name": "csv_aggregate",
"description": "Group a CSV by a column and aggregate a numeric column (sum/avg/count/max/min).",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D7-aggregate-compare",
"world": "DinSteel Solutions",
"industry": "real estate",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S6-D2 | tool | [
"search",
"read",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D1-search-read",
"world": "JaiFoods Industries",
"industry": "real estate",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S59-N9 | tool | [
"data",
"filter",
"negative",
"negative"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "N2-filter-empty",
"world": "ItoCraft Partners",
"industry": "logistics",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S84-D14 | tool | [
"search",
"read",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D1-search-read",
"world": "FonGold Trading",
"industry": "retail",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S116-N3 | tool | [
"data",
"filter",
"negative",
"negative"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "N2-filter-empty",
"world": "SanCare Partners",
"industry": "retail",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S41-D7 | tool | [
"read",
"sections",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_sections",
"description": "List the section headings of a document (table of contents).",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": nul... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D6-sections-read",
"world": "EcoPrime Manufacturing",
"industry": "manufacturing",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S9-D15 | tool | [
"data",
"aggregate",
"multi"
] | [
{
"type": "function",
"function": {
"name": "csv_head",
"description": "Show the first N rows of a CSV table.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": null,
"op": null,
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D3-head-aggregate",
"world": "SomWear Enterprises",
"industry": "food & beverage",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S151-S16 | tool | [
"data",
"filter"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-filter",
"world": "QuinParts Enterprises",
"industry": "logistics",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S90-D13 | tool | [
"search"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-search",
"world": "ChaiSwift Solutions",
"industry": "professional services",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S26-D9 | tool | [
"search"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-search",
"world": "CafeSwift Co., Ltd.",
"industry": "logistics",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S34-D8 | tool | [
"search",
"summarize",
"multi"
] | [
{
"type": "function",
"function": {
"name": "doc_search",
"description": "Full-text search across the document corpus. Returns ranked hits with snippets.",
"parameters": {
"type": "object",
"properties": {
"path": null,
"column": null,
"op": nu... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 2,
"n_turns": 3,
"chain": "D2-search-summarize",
"world": "ItoSupply Partners",
"industry": "professional services",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
S63-S17 | tool | [
"data",
"filter"
] | [
{
"type": "function",
"function": {
"name": "csv_filter",
"description": "Filter CSV rows where a column matches a value or threshold.",
"parameters": {
"type": "object",
"properties": {
"path": {
"type": "string"
},
"column": {
... | [
{
"role": "system",
"content": "You are an expert business information manager for small and medium enterprises. You reason step by step before every action, call only the tools that fit the evidence in front of you, and ground every number in what the document or table actually says. You never invent a fig... | {
"n_tool_calls": 1,
"n_turns": 2,
"chain": "single-filter",
"world": "TepPrime Public Company",
"industry": "retail",
"family": null
} | <|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.
You are an expert business information manager for small and medium enterprises. You reason ... |
SME-SFT-DocData (Qwen3.8-27B Native) — v2.0 production (9,342)
Agentic document + data management SFT dataset for Qwen3.8-27B
(abliterated or base). Teaches the model to manage business documents and data
for small and medium enterprises: search a document corpus, read and summarize
docs, browse folders, and run real aggregations / filters / stats over CSV tables —
then ground every answer in what the tool actually returned. Companion to
hotdogs/cyber-sft-agent-qwen38 (the security tool layer); this v2.0 adds
RAG rows (passive retrieval) and advanced rows (error recovery, multi-doc
synthesis, temporal reasoning, conflict resolution) for production-grade robustness.
v2.0 composition
- 9,342 rows — train 7,799 / valid 1,028 / test 515 (stratified)
- 2,520 tool rows — 11 tools, 16 chain archetypes, every tool output computed from a deterministic per-seed world (a fictional SME's real folder tree + CSVs) so the evidence is consistent, not hallucinated
- 5,880 QA rows — pure document/data-management QA across 5 families (define / process / why / compare / situation), parameterized over 6 SME industries
- 192 RAG rows — passive retrieval: context provided inline (no tool calls), teaches model to reason from retrieved context directly
- 750 advanced rows — error recovery (200), multi-doc synthesis (300), temporal reasoning (150), conflict resolution (100)
11 tools
doc_search · doc_read · doc_sections · doc_summarize · doc_metadata · folder_list · csv_head · csv_aggregate · csv_filter · csv_stats · doc_compare
Document tools: doc_search, doc_read, doc_sections, doc_summarize, doc_metadata, folder_list.
Data tools: csv_head, csv_aggregate, csv_filter, csv_stats. Cross-doc: doc_compare.
Chain archetypes (2,520 tool rows)
| chain | rows |
|---|---|
single-search |
173 |
single-read |
170 |
single-aggregate |
189 |
single-filter |
186 |
D1-search-read |
184 |
D2-search-summarize |
121 |
D3-head-aggregate |
172 |
D4-filter-stats |
189 |
D5-folder-metadata |
156 |
D6-sections-read |
197 |
D7-aggregate-compare |
195 |
D8-compare |
178 |
D9-cross-check |
109 |
N1-search-empty |
9 |
N2-filter-empty |
114 |
N3-read-gap |
178 |
- Single-tool — one well-chosen call for a focused ask (find a doc, read it, roll up a column, filter a table).
- D1-D9 multi-tool — the real SME workflows: search→read, search→summarize, head→aggregate, filter→stats, folder→metadata, sections→read, two-aggregate compare, doc compare, and D9 cross-check (a report makes a claim; the model verifies it against the raw CSV and says whether they agree).
- N1-N3 negative evidence — anti-hallucination: an honest "not found" (N1), an empty filter that is correct because the threshold is above the max (N2), and a doc that exists but doesn't answer the question (N3).
Quality gates (all passing)
| gate | value | target | status |
|---|---|---|---|
| rows | 9,342 | — | — |
| tool-calling rows | 2,520 (27.0%) | — | ✅ |
| RAG rows (passive retrieval) | 192 (2.0%) | >0 | ✅ |
| advanced rows (err/multi/temporal/conflict) | 750 (8.0%) | >0 | ✅ |
| QA rows | 5,880 (62.9%) | — | ✅ |
| multi-turn chains (≥2 calls) | 52.5% | ≥10% | ✅ |
| think-tag coverage | 100.0% | 100% | ✅ |
| tool coverage | 100.0% (11/11 tools) | 100% | ✅ |
| evidence chain (anti-boilerplate) | 82.7% | ≥60% | ✅ |
| duplicate rows | 0.0% | <1% | ✅ |
literal-\n in text |
0.0% | <5% | ✅ |
| negative-evidence rows | 11.9% of tool (301) | >0 | ✅ |
| distractor-palette rows | 41.7% of tool (1051) | ~38% | ✅ |
| token length p50 / p95 / max | 273 / 1,343 / 3,729 | <4,096 | ✅ |
Byte-exact: all 9,342 rows re-render through the real Qwen3.8-27B chat template and verified 9,342/9,342 = 100.0%, 0 render errors. Token length measured with the real tokenizer (Qwen/Qwen3.5-0.8B (vocab 248044)).
Format (the important part)
Rendered through the real Qwen3.8-27B template and verified byte-exact. Format facts (verified via byte ordinals):
| fact | value |
|---|---|
| Role tokens | ASCII: |im_start| / |im_end| wrapped in <> |
| think OPEN | think — 7 chars, NO slash |
| think CLOSE | think/ — 8 chars, HAS slash (asymmetric, don't normalize) |
| tool call | XML: tool_call + function=NAME + parameter=KEY (value on its own line) |
| tool result | role tool → rendered under a user wrapper + tool_response |
| arguments | a dict (template iterates .items()), non-null params in a fixed per-tool order |
Row schema
id— stable row idkind—tool(agentic chain),qa(pure QA),rag(passive retrieval),error_recovery,multi_doc,temporal,conflicttags— e.g.search,data,aggregate,multi,negative,qa,rag,error,temporaltools— function schemas exposed to the model (used tools + optional distractors; empty for RAG/advanced rows)messages— template-native list, rolessystem/user/assistant/tool; thinking inreasoning_content, calls intool_calls: [{name, arguments:{dict}}]meta—n_tool_calls,n_turns,chain,world,industrytext— the full pre-rendered template string (use for packing / single-text SFT)
Splits (stratified)
| split | rows | tool | RAG | advanced | QA |
|---|---|---|---|---|---|
| train | 7,799 | 2,142 | 134 | 525 | 4,998 |
| valid | 1,028 | 252 | 38 | 150 | 588 |
| test | 515 | 126 | 20 | 75 | 294 |
Usage
A. messages[] (trainer calls apply_chat_template)
from datasets import load_dataset
ds = load_dataset("hotdogs/sme-sft-docdata-qwen38", split="train")
row = ds[0]
messages, tools = row["messages"], row["tools"]
B. text (packing / single-text SFT, no template needed — recommended)
text = row["text"] # already byte-exact through the Qwen3.8-27B template
Recommended training config (Qwen3.8-27B, 4× RTX 3060 12GB)
lora:
r: 64
alpha: 128
target_modules: [q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj]
training:
seqlen: 4096 # covers 100% (real max 3,729 tok)
batch_size: 4
gradient_accumulation: 2
learning_rate: 2e-4
epochs: 3
Provenance
Generated by the sme-sft pipeline (Hermes Agent). Tool rows are produced from
deterministic per-seed worlds — each world is a fictional SME with a real folder
tree of documents (contracts, invoices, reports, HR policies, meeting minutes, product
sheets) plus CSV tables (sales, invoices, employees, expenses, customers); every tool
output is computed from that world, so a summary matches the doc and an aggregate
matches the CSV. QA rows are drawn from a combinatorial generator over 6 SME
industries and 5 question families. Date anchor: 2026-08-20. Format ground truth:
byte-ordinal analysis of the real template, validated 100% against
hotdogs/cyber-sft-agent-qwen38.
- Downloads last month
- 26