claim
active
claim:appeasing-untrained-test-biases-indicates-a-general-rm-pleasing-strategy-rather-than-rote-learningAppeasing untrained test biases indicates a general RM-pleasing strategy rather than rote learning.
Summarizing Marks et al.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Behavioral baseline.
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.
- Central threat model claim derived from RL experimental results
- Interpretive claim explaining why tuned models fail neutral and low-valence personas
- Authors' interpretation of prompt variation results showing alignment faking disappears only when conflicting objective is removed
- Training that suppresses representations of undesirable concepts may incentivize new representations, removing the detection signalhypothesis0.764Tension for monitorability.
- Mechanistic explanation for the increase in AF reasoning during RL
- Novelty claim.
- Practical application proposed based on mechanistic findings