finding
active
finding:aos-match-or-exceed-best-prior-method-on-3-of-4-auditing-tasksAOs match or exceed best prior method on 3 of 4 auditing tasks
Headline auditing result.
Source paper
extracted_from(2025) · Adam Karvonen
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Activation Oracles generalize far beyond their training distributionassociated_withsupportsCentral claim from tl;dr.
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.
- Scaling result over Qwen-3 8B, Gemma-2 9B, Llama-3.3 70B, Claude Haiku 3.5.
- Limitation relative to SAEs.
- Downstream task validating NLA utility for model auditing; agents succeed without access to misalignment training data.
- Injecting the activation difference between fine-tuned and original model into the AO.
- Appendix C.
- Overall positioning.
- Cross-test agreement.
- Argument for scalability.