method
active
method:revealed-preferences-evaluationRevealed Preferences Evaluation
Novel evaluation method that measures a model's preference to express one character trait over another via Elo scoring, avoiding self-report issues
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.
- Behavioral and stated consistency that implies the model is pursuing some objective, without claiming genuine internal states
- The ability of active inference agents to learn their own prior preferences over outcomes by accumulating Dirichlet parameters from experience.
- The problematic possibility of digital minds with superhumanly strong preferences requiring interpersonal utility comparison frameworks
- Target distribution over states or outcomes encoded in the generative model; goal states.
- Post-training alignment method during which undesirable behaviors emerged in the studied model.
- Replaces explicit reward signal in active inference; encodes agent's preferred observations independent of environment.
- Designing digital minds to have preferences that are trivially easy to satisfy, yielding high welfare at minimal resource cost
- Key element for alignment faking: model's pre-existing preferences contradict the new training objective