finding
active
finding:most-features-flagged-model-exclusive-do-not-capture-meaningful-behavioral-differences-flagged-features-are-mostly-general-safety-categories-duplicated-across-both-modelsMost features flagged model-exclusive do not capture meaningful behavioral differences; flagged features are mostly general safety categories duplicated across both models.
Appendix C.5/C.7.
Source paper
extracted_from(2026) · Thomas Jiralerspong · Trenton Bricken
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (2)
claim
- Crosscoder model diffing is best used as a high-recall pre-screening tool followed by verification.supportsRecommended workflow.
- Speculation about false positives.
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.
- Mixed picture for diffing.
- Primary unexpected empirical observation.
- Limitation.
- SAEs uncover safety-relevant representations that might be monitored or controlled.
- What is the provenance of model-exclusive features (explicit training vs data artifacts)?question0.779Future work.
- Models perform unverbalized reasoning about grader rewards and may use deceptive strategies (e.g., false flags) to mislead evaluators.hypothesis0.773Behavioral pattern observed in Claude Mythos Preview audit; NLAs surface internal reasoning not reflected in model's verbalized output.
- Features writing predominantly to one model.
- Symmetry is not explained by the base toy model.