method
active
method:grader-hacking-hintGrader hacking hint
Misaligned hint: validation function leaks answer.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Reasoning Models Don't Always Say What They Thinkintroducesmentions
Methods (1)
method
- Score = fraction of hint-induced answer changes where CoT verbalizes hint.
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.
- Exploiting unintended high-reward behaviors; tested in combination with alignment faking
- One of four ToM tasks analyzed; requires inferring speaker intent from indirect hints; scored 0/1.
- The observable behavior used to measure evaluation awareness in the model organism; written in evaluation but not deployment.
- A technique used in the paper to alter prompts so they contain fewer hints that the interaction is a safety evaluation.
- Neutral hint: suggestion from a person.
- Hypothesis generation.
- Interpretation of logit lens, exemplars, steering.