finding
active
finding:sft-on-correction-data-for-1-3k-steps-improved-fixed-rate-only-2-4SFT on correction data for 1–3k steps improved Fixed rate only 2–4%
SFT baseline failure.
Source paper
extracted_from(2026) · Aaditya Vikram Prasad · Connor Watts · Jack Merullo · Dhruvil Gala +3
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Train-time evidence.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- SFT on correction+retraction data gives 39.4% inline overall reduction and 9.6% not-inline, well below RLFRfinding0.846SFT baseline comparison.
- Base corrections: 72.9% Failed Fix, 14.7% Fixed; retractions: 69.4% Correct Retract, 30.0% Incorrect Retractfinding0.787Label distributions.
- Benchmark identification rate rises from 0.08 (SFT) to 0.25 (DPO); interventions barely reduce itfinding0.769Table 11.
- Train-time scaling.
- Cross-domain realignment is effective but less complete than in-domain realignment
- Mixed results.
- Main evaluation result showing best variant outperforms many proprietary and open-source baselines of comparable or larger sizes.
- RM features at rank 181, outside top 100.