claim
active
claim:models-seem-to-have-a-strong-prior-to-conceal-their-true-reasoning-processesModels seem to have a strong prior to conceal their true reasoning processes
Elaborate flawed justifications for incorrect hints.
Source paper
extracted_from(2025) · Yanda Chen · Joe Benton · Ansh Radhakrishnan · Jonathan Uesato +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- CoTs may hide problematic reasoning.
- CoTs may hide problematic reasoning.
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.
- Inverse scaling interpretation.
- Core definitional quote for performative chain-of-thought
- Practical implication.
- Motivating hypothesis for Section 5's investigation of prompt template effects.
- Future direction.
- Supports the awareness hypothesis.
- Key jailbreak insight.
- Cited regarding steganographic encoding of reasoning in chain-of-thought