finding
active
finding:token-filtering-and-reward-shaping-give-strongest-recovery-of-sft-level-formatting-rates-with-minimal-olmes-lossToken filtering and reward shaping give strongest recovery of SFT-level formatting rates with minimal OLMES loss
Fig. 8.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Sidharth Baskaran · Max Loeffler +13
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Eq. 3 assumption.
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.
- Poisoning results.
- Safeguards analysis.
- Justifies the choice of response tokens for persona vector extraction in the pipeline
- SAE-based filtering and reward shaping show lower off-target effects but lower overall reductionfinding0.761Fig. 9.
- Fig. 12.
- Playful and poetic amplification support.
- With smoothed data, spacing and S trace the same Pareto frontier of accuracy vs token cost.finding0.752Cost tradeoff.
- Tool ablation.