claim
active
claim:activation-probes-perform-similarly-to-llm-monitors-at-catching-reward-hacking-at-far-lower-costActivation probes perform similarly to LLM monitors at catching reward hacking at far lower cost
Practical claim.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Andrew Lee · Barak Widawsky +14
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (4)
finding
- DoM probes catch 3.1% more hacks in Kimi K3 but 7.9% fewer in GLM 5.2 on DeepSWE at matched FPRsupportsProbe vs LLM monitor performance.
- Cost of replacement.
- Per-environment comparison.
- Monitor replacement.
Claims (1)
claim
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.
- Discovery claim.
- Joint use improves TPR.
- Interpretation of resampling.
- Discussion positioning.
- Further studying the relation between activation probes and the logit monitor could help better understand activation probeshypothesis0.788Future work.
- Open scientific question motivating the work.
- Fixed and Retracted rates are essentially unchanged when reward probe runs on base versus policy activationsfinding0.787Reward pipeline transfer.
- Justifies restricting probe-based vector derivation to h_b activations; attributed to Yes/No semantics