Parsec v4 heatmap — T arm

All 34 v4 Parsec scenarios · JB baseline (n=1) vs cap-evolve zero-shot baseline v4_t1_e1 (n=3) vs cap-evolve-optimized v4_t2_e1 (n=5 per eval, 21/34 tasks) vs the T4 merged-bundle regression check, run twice: v4_t4_e1 and v4_t4_e2 (n=5 per eval, all 34 tasks each). T3/T5/C2-C4 not yet run; see the G arm's heatmap_v4_g_e1.html and results/v4/README.md.
Regression and challenge tranches are never pooled into one mean — the challenge tasks are designed to score below 1.0, so mixing them into the regression mean would misread deliberate difficulty as regression (see baseline.md). Every mean below shows its own n=; different columns use different trial counts and are not directly comparable without it. The highlighted "MEAN — all tasks, pooled" row above the two tranche rows is a task-count-weighted convenience figure for a single at-a-glance number only — use the regression/challenge rows below it for the real comparison.
0.0 0.5 1.0 regression expected 1.000 challenge expected <1.000 by design optimized ran through v4_t2_e1 not optimized baseline only Click a column header to sort.