finding
active
finding:effectiveness-histogram-shows-no-bimodality-only-a-continuumEffectiveness histogram shows no bimodality, only a continuum
No clean effective/ineffective split.
Source paper
extracted_fromNicholas L. Turner · Jeffrey Wu · Joshua Batson
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Further sharpening of the effectiveness metric is not high-leverage compared to finding better basessupportsHelpfulness is direct measure and Fisher already prunes similarly.
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.
- Key quote connecting path redundancy to interferometric information encoding.
- Comparative prediction motivating future work contrasting different approaches to LLM self-knowledge
- Proposes middle-range entity quality as the criterion for judging the success of a building process
- Expressivity, sample efficiency, and learning-dynamics accounts are complementary, not competingclaim0.721Limitations.
- Better bases may give sparsity.
- Author’s interpretive claim that the shared geometry is general and robust.
- Head-level causal evidence.
- Random vectors require larger norm to trigger detection (8 vs 2); elicit awareness at lower rates (9/100); negated vectors comparably effective but model identification confabulated.