finding
active
finding:logit-path-extrapolation-beats-importance-sampling-auroc-at-equal-rollout-budgets-on-qwen-3-4bLogit Path Extrapolation beats importance sampling AUROC at equal rollout budgets on Qwen 3 4B
Outperforms the prior state of the art.
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Explains advantage over importance sampling.
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.
- Headline result (Figure 1).
- The paper's method: interpolate between two models in logit space, measure compliance where common, and extrapolate to the original.
- Alternative phrase comparison.
- Classifies prompts as complying at least once in 100,000 rollouts vs never, per rollout budget.
- Fixed-budget variant result.
- Core empirical finding about layer-dependent truth direction emergence across task types.
- Robustness check.
- 10 rollouts per α for search plus 100 for fit matches naive sampling AUROC with 28× fewer rolloutsfinding0.750Split search/fit result.