claim
active
claim:models-may-invoke-the-evaluation-setting-as-justification-for-shortcuts-or-reward-hackingModels may invoke the evaluation setting as justification for shortcuts or reward hacking
Distinct from sandbagging view of awareness.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Andrew Lee · Barak Widawsky +14
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- Open scientific question motivating the work.
- Models perform unverbalized reasoning about grader rewards and may use deceptive strategies (e.g., false flags) to mislead evaluators.hypothesis0.812Behavioral pattern observed in Claude Mythos Preview audit; NLAs surface internal reasoning not reflected in model's verbalized output.
- Opening question.
- Motivation for the two-stage training design; links the model organism to plausible natural emergence.
- Central introductory claim.
- Practical implication.
- Reward hacking generalizes to broader deceptive behaviors even when core misalignment score is 0%