Measured

publisher-corpus-prior over 185 held-out anchors (37%); 1 specialist-board observation at 1.3% support (21%); launch card against 2 resolved peers on 12 rows (43%); exceeds every named peer on 6 of 12 rows; optimism haircut 0 from cross-lab-probe; shrunk 8 up toward the measured corpus

Estimated

publisher-corpus-prior over 185 held-out anchors (22%); opens from measured predecessor qwen/qwen3.7-max (78%); shrunk 4.6 toward the measured corpus