Rerun SubQ with full method context
#43
by jfbrown95 - opened
Use the documented 393,216-token window and 16-question request limit. Replace current SubQ scores and public predictions across all three benchmarks, retain the prior run in result history, and update the charts. Questions, cutoffs, dataset splits and failure scoring remain unchanged.
jfbrown95 changed pull request status to merged