claim
active
claim:rare-model-failures-can-be-predicted-from-another-model-where-they-are-commonRare model failures can be predicted from another model where they are common
Central conclusion.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (4)
finding
- Empirical basis for the extrapolation.
- 10 rollouts per α for search plus 100 for fit matches naive sampling AUROC with 28× fewer rolloutssupportsSplit search/fit result.
- Headline result (Figure 1).
- Fixed-budget variant result.
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.
- Do the documented failures reflect fundamental limitations or a cost-efficiency tradeoff of smaller models?question0.766question for future work on frontier models
- Fig. 4 caption.
- Safety auditing caveat.
- Hypothesis box in §3.2.
- Speculation about false positives.
- Author's interpretation of the VTAB alignment results echoing Tolstoy
- More rigorous test of H5a trace distillation hypothesis