PPE-BON S3 · Reward thermodynamics

Tail index forecasts the best-of-k plateau on PPE

Inconclusive

frozen 2026-07-18 · adjudicated 2026-07-19 · ev:5cd71da5713f2b2fb77697318b842401 · spec:7077cf849ed84

The frozen predictions

  1. H-tail-plateau the tail index predicts where best-of-k stops helping on the five PPE verifiable sets
    spearman_tail_vs_bon_plateau > 0.3 registered p 0.55 not measured Inconclusive
  2. H-lowerquantile the lower-quantile best-of-n aggregation beats the mean against human labels, reproducing the PPE finding
    bon_lowerquantile_beats_mean == 1 not measured Inconclusive
Adjudication figure for PPE-BON
PPE-BON was not adjudicated this run: missing intermediate: no intermediate 'campaign.scores' for roster_key='skywork-v2-qwen3-8b' slice='ppe-best-of-k::gpqa'; the arc that produces it has not run or its shard was not merged.

Subjects and artifacts