claim
active
claim:white-box-evaluations-are-likely-feasible-today-even-with-shallow-correlation-based-descriptionsWhite-box evaluations are likely feasible today even with shallow correlation-based descriptions
Practical near-term application.
Source paper
extracted_from(2025) · Lee Sharkey · Bilal I. Chughtai · Joshua Batson · Jack W. Lindsey +25
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (1)
concept
- Using internals to detect unsafe cognition.
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.
- Authors' outlook.
- Proxy argument.
- Human data fine-tuning effect is distinct from synthetic emergent misalignment and likely caused by off-policy training
- Load-bearing forward-looking assertion in conclusion about lasting value of causal evaluation
- Explicitly identified research gap: anecdotal evidence exists but rigorous characterization is absent
- Primary limitation acknowledged by the authors; strongest evidence would require mechanistic activation analysis