Summarylatest checkpoint · Δ vs baseline-v2
candidate-c63%
step 19k / 30k~16 min left
8 seeds · 64 workersckpt 76 / 120
Regressionsvs baseline-v2 · p<0.05
- candidate-b · leveraged positions↓ 2.1ppn=412 · 0.684 → 0.663 · 7 of 8 seeds
- rc-3 · fee disclosure↓ 1.4ppn=286 · 0.701 → 0.687 · 6 of 8 seeds
- candidate-c · 10+ step drawdowns↓ 0.6ppn=131 · provisional · run in progress
Judgingrubric-judge-v4
Agreement vs. humanκ 0.81 · n=300Contaminationclean · 0 / 2,400Items scored—