crucible-sft-gemma-4-12b-it-mini

# crucible-sft-gemma-4-12b-it-mini Self-distilled SFT dataset of verified reasoning traces from **unsloth/gemma-4-12b-it**, built by the [reasoning-compression](https://github.com/marcodsn/reasoning-compression) crucible pipeline: k-sample generation on a decontaminated prompt pool, inline verification (symbolic math / sandboxed code tests), difficulty banding via solve rate, and loop-detector filtering on the chosen trace. Each row: `prompt`, `reasoning` (a verified-correct thinking trace when the domain is verifiable), `response`, `domain`, `verified`, `difficulty` (`{solve_rate, k}`), sampling params in `gen`. - Rows: **824** — domains: {"math": 196, "creative": 308, "general": 259, "code": 61} - Prompts are content-addressed (`prompt_id` = sha256 of prompt text) and decontaminated against GSM8K, MATH-500, HumanEval, MBPP, MMLU-Pro. - A stratified held-out split was carved *before* generation and is not included here. Part of the caveman reasoning-compression ablation study (see the flint datasets built from this file and the trained comparison arms). ## Build manifest ```json { "created_at": "2026-07-10T23:23:51+00:00", "git_sha": "9b2ffbbeeddc3f9094805e1ea9a07d2dedaf2eb7", "recipe": "flint-ablation-v1-mini", "recipe_def": { "target_size": 1100, "shares": { "math": 0.3, "code": 0.1, "general": 0.32, "creative": 0.28 }, "candidate_caps": { "math": 700, "code": 250, "general": 400, "creative": 350 }, "solve_band": [ 0.2, 0.8 ], "k": 4, "oversample": 2.2, "lang": "en" }, "raw_rows": 1620, "selected_rows": 824, "domains": { "math": 196, "creative": 308, "general": 259, "code": 61 }, "verified": { "True": 257, "None": 567 }, "degenerate_in_raw": 86, "gen_model": "unsloth/gemma-4-12b-it" } ```