finding
active
finding:explain-away-steering-on-rejected-responses-outperforms-chosen-response-steering-tulu-70b-safety-0-864-vs-0-779Explain-away steering on rejected responses outperforms chosen-response steering (Tulu 70B safety 0.864 vs 0.779)
Table 10.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Sidharth Baskaran · Max Loeffler +13
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (2)
claim
- Model may fight chosen-response steering.
- Contrast with layer-specific steering.
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.
- Fig. 10, Table 10.
- Practical disadvantage of activation steering highlighted as a drawback
- Demonstrates distributed steering is more effective and less accuracy-damaging than concentrated steering.
- Models trained with amplification forward steering show reduced efficacy of refusal steering vectorsfinding0.760Fig. 10c.
- Illustrative finding that ESR mitigates but does not fully eliminate steering influence
- Practical finding for optimizing steering setup.
- Second significant cross-concept introspection improvement; marginal after BH correction (q≈0.066)
- Demonstrates averaging multiple prompt pairs reduces noise; optimal subset selection further improves performance.