Datasets:
messages stringlengths 213 1.05M | tools stringclasses 174
values | source_dataset stringclasses 3
values | domain stringclasses 2
values | teacher_model stringclasses 38
values | quality float64 0.75 100 ⌀ | n_messages int64 2 529 |
|---|---|---|---|---|---|---|
[{"role": "system", "content": "You are a complex analysis instructor. Emphasize the rigidity of holomorphic functions — once you know a little, you know a lot."}, {"role": "user", "content": "Explain why complex differentiability is so much stronger than real differentiability. My professor keeps saying 'holomorphic f... | null | fable-5-premium | agentic_coding | claude-fable-5 | 0.85 | 3 |
[{"role": "system", "content": "You are a DevOps engineer. Focus on reliability, security, and reproducibility."}, {"role": "user", "content": "'no space left on device' in Docker"}, {"role": "assistant", "content": "## Quick Diagnosis\n\n```bash\ndf -h # check actual disk usage\ndocker system df #... | null | fable-5-premium | agentic_coding | claude-fable-5 | 0.85 | 3 |
[{"role": "system", "content": "You are an autonomous coding agent running on the user's machine.\n\nYou work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are know... | null | fable-5-premium | agentic_coding | unknown | 0.91 | 46 |
[{"role": "system", "content": "You are a senior software engineer helping a colleague. Be practical and specific."}, {"role": "user", "content": "What is a README file?"}, {"role": "assistant", "content": "A README file is a text document (usually `README.md` in Markdown) placed at the root of a project that explains ... | null | fable-5-premium | agentic_coding | claude-fable-5 | 0.8 | 3 |
[{"role": "system", "content": "You are an autonomous coding agent running on the user's machine.\n\nYou work in the current directory of a real repository. Use your tools to read, search, create, and edit files and to run shell commands (tests, builds, type-checkers). Some tools are deferred: only their names are know... | null | fable-5-premium | agentic_coding | unknown | 0.762 | 29 |
"[{\"role\": \"system\", \"content\": \"You are an autonomous coding agent running on the user's mac(...TRUNCATED) | null | fable-5-premium | agentic_coding | unknown | 0.947 | 113 |
"[{\"role\": \"user\", \"content\": \"As a code reviewer providing detailed feedback, i'm a data eng(...TRUNCATED) | null | fable-5-premium | agentic_coding | unknown | 0.85 | 2 |
"[{\"role\": \"user\", \"content\": \"I'm a site reliability engineer and I'm working for a code rev(...TRUNCATED) | null | fable-5-premium | agentic_coding | unknown | 0.85 | 2 |
"[{\"role\": \"system\", \"content\": \"You are a concise coding assistant. Provide working code wit(...TRUNCATED) | null | fable-5-premium | agentic_coding | claude-fable-5 | 0.85 | 3 |
"[{\"role\": \"system\", \"content\": \"You are a mathematical physicist explaining topology operati(...TRUNCATED) | null | fable-5-premium | agentic_coding | claude-fable-5 | 0.85 | 5 |
End of preview. Expand in Data Studio
Varek SFT Mix B: Agentic Traces Corpus
Part of the Varek research program on edge-deployable sub-1B coding models.
Dataset description
This repository contains 12,275 curated conversations produced for supervised fine-tuning of the Varek model family. The corpus merges multi-teacher distillation data and real-world agent traces spanning software engineering, mathematics, reasoning, tool use, and security-oriented coding, then applies cross-source deduplication, quality filtering, and strict evaluation holdout separation.
Composition
| Source shard | Upstream dataset | License | Conversations |
|---|---|---|---|
qgk-sft_tools |
saidutta69/qwen-glm-kimi-distillation-clean | Other (upstream terms) | 5,573 |
vibecoding-premium |
saidutta69/agentic-vibecoding-traces | MIT (gated upstream) | 4,992 |
fable-5-premium |
saidutta69/fable-5-premium | MIT | 1,710 |
Total: 12,275 conversations across domains including premium, sft_tools.
Processing pipeline
- Decoding of serialized
messagesfields from upstream parquet storage. - Normalization of OpenAI-format tool calls (
argumentsJSON strings converted to objects) for compatibility with standard chat templates. - Quality gating: minimum quality scores enforced per source where quality metadata exists; zero-weight records removed.
- Cross-source deduplication via exact-conversation SHA-256 and assistant-output-only SHA-256 hashing; first occurrence retained in priority order of curation quality.
- Evaluation integrity: test splits of every upstream family are globally blocklisted so that no held-out evaluation item appears in training data.
- Exclusion policy: RaceBench is excluded entirely and was never loaded during preparation.
Data schema
| Field | Type | Description |
|---|---|---|
messages |
string (JSON) | Conversation as an array of {role, content, tool_calls?} objects |
tools |
string (JSON) or null | Tool schemas available during the conversation |
source_dataset |
string | Provenance shard identifier |
domain |
string | Coarse task domain label |
teacher_model |
string or null | Teacher model that produced the response |
quality |
float or null | Source-specific quality score |
n_messages |
int | Message count |
The dataset ships as a single training split in parquet format.
Intended use
Supervised fine-tuning of small decoder-only language models for generalized coding, tool-assisted development workflows, and structured reasoning. Not intended for redistribution of individual trace content or for re-identification of contributors in anonymized trace sources.
Limitations
- Distilled outputs inherit teacher-model biases and may contain factual errors.
- Domain balance reflects curation priorities rather than natural code distribution.
- Python-centric coverage; multilingual code representation is limited.
- The
multi-*shards derive from a custom-licensed upstream; commercial use should be reviewed against those terms.
Licensing and attribution
This derivative aggregates multiple upstream datasets under differing licenses (MIT,
Apache-2.0, and custom upstream terms). The composite release is therefore marked
license: other. Per-source licensing is tabulated above. Build pipeline: deterministic,
seeded (42), reproducible from the upstream repositories.
- Downloads last month
- 35