claim
active
claim:training-helps-most-where-the-mechanism-behind-the-behavior-is-vagueTraining helps most where the mechanism behind the behavior is vague
From unfiltered results.
Source paper
extracted_from(2026) · Adam Karvonen · Euan Ong · Subhash Kantamneni · Samuel Marks
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Gains concentrated in low-concreteness claims.
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.
- Training that suppresses representations of undesirable concepts may incentivize new representations, removing the detection signalhypothesis0.779Tension for monitorability.
- Developmental analogy used to explain sample efficiency under high ρd conditions
- Confident wrong counterfactuals mislead.
- Core goal statement.
- Claim that models learn the spirit of the constitution, not just its letter, evidenced by suppression of opposing traits
- If the relationship is present in activations, better tools (e.g. trained on these datasets) could surface ithypothesis0.772Alternative explanation.
- Foundational motivation for the research.
- Central interpretive claim and motivation for future work