Upload README.md with huggingface_hub

README.md

@@ -25,7 +25,7 @@ diversity (NJD), computed via `kin_flow.cli.bench` on 10 held-out test scenes wi
 | `se-shadow_5000_40` | single-embodiment | Shadow Hand (22 DOF) | 40 / 500 | 75.9% | 0.232 | RTX 6000 Ada |
 | `me-full_25000_5` | multi-embodiment | all 5 + z0 | 5 / 500 | 82.1% (mean) | 0.212 (mean) | MI300X (ROCm) |
 | `me-full_25000_30` | multi-embodiment | all 5 + z0 | 30 / 500 | 84.3% (mean) | 0.203 (mean) | MI300X (ROCm) |
-| `me-full_25000_90` | multi-embodiment | all 5 + z0 | 90 / 500 | eval pending | eval pending | MI300X (ROCm) |
+| `me-full_25000_90` | multi-embodiment | all 5 + z0 | 90 / 500 | 86.4% (mean) | 0.212 (mean) | MI300X (ROCm) |
 
 Single-embodiment models: `num_scenes=5000`. Multi-embodiment model: the paper's full
 configuration — `num_scenes=25000`, fp32, batch 5 scenes × 128 grasps, warmup-cosine LR
@@ -55,6 +55,17 @@ step; epoch numbers here are the corrected true dataset-pass count.
 | Shadow Hand | 22 | 71.6% | 0.224 |
 | **mean** | | **84.3%** | **0.203** |
 
+### `me-full_25000_90` per-gripper eval (epoch 90)
+
+| Gripper | DOF | SR | NJD |
+|---|---|---|---|
+| Panda | 2 | 96.8% | 0.265 |
+| VX300 | 2 | 96.7% | 0.155 |
+| DexEE | 12 | 73.1% | 0.109 |
+| Allegro | 16 | 86.4% | 0.289 |
+| Shadow Hand | 22 | 79.2% | 0.243 |
+| **mean** | | **86.4%** | **0.212** |
+
 Approaching convergence (epoch 90 of ~120 needed, per the original repo's own README
 guidance) — newer-epoch checkpoints will be added as training progresses.