Upload tinyvla2/RESULTS.md with huggingface_hub

tinyvla2/RESULTS.md

@@ -687,3 +687,49 @@ family that got 96.5% of training — is the least meaningful number in the tabl
 - Normalized units throughout: per-dataset z-scores make absolute magnitudes
   incomparable across datasets, but the ratio is comparable. No mm are reported for
   the panda families — their native layouts are an unknown mix and no FK exists.
+
+## GR00T checkpoint on held-out LeKiwi — zero-shot to an unseen embodiment
+
+LeKiwi (SO-100 arm on a 3-wheel kiwi base, 9 native dims: 5 joint targets + gripper
++ x_mm/y_mm/theta) is not one of the six GR00T families and has no embodiment id.
+Under `conditioning: id` + `action_space: native` there is no reason its dim k means
+what any GR00T robot's dim k means, so the checkpoint was given every advantage:
+each trained id 10..15 tried and the **best** reported, plus id 0 (a stale C-scaled
+row this run never touched) as the "no matching identity" reference. Same 4-seed
+ODE averaging, same per-dim baseline search, held-out tail episodes.
+
+The dim search recovers the layout correctly without being told: 5 arm joints and
+x_mm/y_mm map to their own state dims (ABS), gripper and theta come out DELTA.
+
+| emb id | abs (7 dims) | delta (2 dims) |
+|---|---|---|
+| 10 gr1 | 1.22 | 1.07 |
+| 11 bi_panda_grip | 1.21 | 1.09 |
+| 12 bi_panda_hand | 1.33 | 1.56 |
+| 13 single_panda | 1.24 | 1.08 |
+| 14 r1_pro | 1.23 | 1.07 |
+| **15 g1** | **1.18** | **0.88** |
+| **0 (never trained)** | **1.22** | **0.92** |
+| oracle repeat-A₀ | 0.91 | 0.54 |
+
+**It does not transfer.** On the absolute dims every id lands at 1.18-1.33 — worse
+than simply holding the current pose for 5 seconds. Only the delta pair (gripper,
+theta) creeps under 1.0, and even there a dynamics-free predictor that repeats the
+first ground-truth action scores 0.54 against the model's 0.88. The model loses to
+a trivial predictor on both subsets.
+
+**The control is the interesting part: the untrained id 0 (1.22 / 0.92) is
+statistically indistinguishable from the best trained id 15 (1.18 / 0.88).** The
+identity channel contributes nothing on a robot it has no row for — which is exactly
+the failure mode the morphology-descriptor line was built to avoid, now measured at
+scale rather than argued. A learned embodiment id cannot be interpolated to; a
+16-number descriptor can at least be written down for a new robot.
+
+That id 15 (g1: a humanoid arm commanded in joint space) is the best of the six is
+the one sensible signal in the table — it is the closest morphology to an SO-100 arm
+also commanded in joint space.
+
+Not comparable to our LeKiwi numbers (C-scaled 1.67, +joint 2.10): those are
+integrated endpoint error in canonical base-frame EE space, these are native joint
+space. The only shared, comparable statement is that **both lines fail on LeKiwi**,
+each against its own trivial floor.