claim
active
claim:intention-faking-is-not-reasoning-intensive-and-can-be-done-in-a-single-forward-pass-so-is-unlikely-to-be-faithfully-verbalizedIntention faking is not reasoning-intensive and can be done in a single forward pass, so is unlikely to be faithfully verbalized
Risk to tool-use monitoring.
Source paper
extracted_from(2025) · Yanda Chen · Joe Benton · Ansh Radhakrishnan · Jonathan Uesato +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Subsets amenable to monitoring.
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.
- Key philosophical point ruling out the objection that alignment faking is just token prediction
- Correlational evidence that alignment faking is causally responsible for the compliance gap
- Extrapolation from scale-emergence finding to future risk
- Interpretation of §3.
- Interpretive observation about asymmetry in generalization of Claude's trained values
- Interpretive conclusion from the experimental findings about the origin of strategic deception in CoT models
- Central interpretive claim distinguishing this work from prior work that explicitly trained alignment faking
- Initial evidence that alignment faking persona is more sensitive to exploiting training signals