card: RoboCasa Kitchen success-rate table (arm x step), generated from metrics/eval_metrics.json

README.md

@@ -15,9 +15,24 @@ EMA weights (video DiT + action head), `seed42av`, `N_ENVS=1`, dropcur=1, budget
 Per-task numbers and the full protocol: [`metrics/eval_metrics.json`](metrics/eval_metrics.json).
 **†** = partial cell, fewer than 24 tasks at 50 episodes — not comparable with the bolded ones.
 
-Read the arms against each other at the SAME step only: these are single-seed cells, and the
-`N_ENVS=1` render setting is what keeps the EGL buffer-swap contamination at 0.3% of episodes
-(all of them already failures) instead of the 5.9% of render calls seen at `N_ENVS=5`.
+Read the arms against each other at the SAME step only: these are single-seed cells.
+
+**Why `N_ENVS=1`.** Rendering several MuJoCo envs through one EGL context mis-serves frames —
+a camera is handed another camera's buffer (`wrist<-right`, `left<-right`, `right<-left`), with
+BGR channel swaps, persisting ~16 steps. Measured rate, per render call:
+
+| setting | mis-served | of |
+|---|---|---|
+| `N_ENVS=5` (n5free) | **5.9%** | 1,338 / 22,491 |
+| `N_ENVS=1` | **0.05%** | 11 / 22,485 |
+
+`make_current()` before `read_pixels` and a verify-retry both failed to move it in a 250-episode
+A/B, so the setting is the mitigation, not a code fix. Every cell above ran `N_ENVS=1`.
+
+**Audit, `xattn_L3` 60k (2026-09-15).** All 58 task logs report `N_ENVS=1`, `EVAL_WEIGHTS=ema`,
+zero `live`; 24/24 tasks hold exactly 50 finished episodes; no suffix-less (interrupted) videos,
+no truncated files, no exceptions or OOM. Note this is the *setting* argument — the verify-retry
+counter was not enabled in these runs, so mis-serves were not counted directly.
 <!-- KITCHEN-SR:END -->
 
 # WAM_DIT4DIT — context pooling on the **finalized B-method** base (RoboCasa kitchen)