hypothesis
active
hypothesis:humans-prefer-representative-distributions-over-non-representative-ones-when-evaluating-outputs-at-the-distribution-level-providing-a-reward-gap-that-vs-exploitsHumans prefer representative distributions over non-representative ones when evaluating outputs at the distribution level, providing a reward gap that VS exploits
Assumption D.3 formalized in the theoretical framework; empirically validated with coin-flip sequence experiments
Source paper
extracted_from(2025) · Jiayi Zhang · Simon C.H. Yu · Derek Chong · Anthony Sicilia +3
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.
- Core framing.
- Abstract; central distinction.
- Claim about broader applicability of the scaling argument
- Playful and poetic amplification support.
- Justification for reward probes.
- The theoretical mechanism explaining why VS works despite mode collapse remaining operative
- Distinct from sandbagging view of awareness.
- Base models assign higher likelihood to typical-set (representative) sequences than to degenerate sequences under VS promptshypothesis0.768Assumption D.6 formalized in the theoretical framework; empirically validated with coin-flip typicality rating experiments