# SFT4B: SFT4B

> Living preprint · as of 2026-10-25 · map node X-153 (experiment) · interim record, not a final result.

<!-- lint:exempt:start -->
- [X-153] SFT4B · SFT4B · question: Can a small LoRA (rank 16, ~1,200 extraction-style grounded QA examples from the packs' own text, filtered by jev quote_supports >= 0.9) teach the claim 4B reader to use the kept pack passage, measured as X (claim 4B + LoRA) vs C (claim 4B, SELFTEST-SCORES with-pack traces) on the 180 pack-matched self-tests, with v0.1 public 60 no-harm. · result: pass 38 %, 562 kept vs ≥ 1,000. · set 180 self-tests; v0.1, judge none (deterministic gate), n 180 (eval, planned); 1,200 train / 100 hold-out examples built, seeds 1 (deterministic generation; training seed 42) · verdict: closed (open / in progress: stage 0 data PASS (v1, then v2); stage 1 environment + parity tooling PASS on the smoke adapter (STAGE1.md); training (step 6) on hold until v3 data (addendum 2); no verdict)

<!-- lint:exempt:end -->
Read chapter 9, Knowledge packs: https://r4topunk.github.io/boar-research/research/packs.md

## Facts cited {#facts}

Resolve every [F-…] marker in https://r4topunk.github.io/boar-research/data/facts.json. Each line gives the claim, the number with its metric family, set and status, judge, n, seeds, CI, date and caveat.

<!-- lint:exempt:start -->
- [F-508] SFT4B: LoRA fine-tune of the 4B to read kept passages → teacher faithfulness 38 % (Haiku), 5 % (deep) vs 50 % (no faithful teacher)
- [F-549] R71 Grounded-reading SFT with Haiku teacher: data gates. pass 38 %, 562 kept vs ≥ 1,000 · set UNKNOWN (status UNKNOWN) · judge UNKNOWN · n UNKNOWN · seeds UNKNOWN · CI not computed · caveat: why closed: stage-0 FAIL
- [F-640] Reader-side fine-tuning teacher faithfulness (SFT4B) Haiku teacher 38–45 %; deep-tier teacher 6/120 (5 %) vs 50 % bar · set distractor-rich engine renderings (status UNKNOWN) · judge jev · n UNKNOWN · seeds UNKNOWN · CI 120 contexts · date 2026-10-13 22:28 / 10-14 10:56 · caveat: 45/120 deep answers empty (shared token budget)

<!-- lint:exempt:end -->
---

Cite as: r4to, independent researcher. RatLab — A Boar Research (2026), research record 1.0, as of 2026-10-25. Text and data CC BY 4.0. Fact ids or a link must travel with a quote. How to cite: https://r4topunk.github.io/boar-research/cite/ · https://r4topunk.github.io/boar-research/CITATION.cff.

AI-reviewed, not clinically reviewed. Nothing here is medical advice.