Does VLA-JEPA run on Franka Emika Panda?
partial
Scope: Simulation only: LIBERO (simulated Franka Emika Panda); no real-robot training or inference on a Panda is documented. The paper's real-robot experiments use a Franka Research 3 (see the franka-fr3 cell).
Capabilities
| Capability | State | Notes | Evidence |
|---|---|---|---|
| Train / fine-tune | no verified evidence | — | |
| Real-robot inference | no verified evidence | — | |
| Simulation inference | works | LIBERO (four suites) in the paper; LeRobot v0.6.1 documents fine-tuning on HuggingFaceVLA/libero and evaluating lerobot/VLA-JEPA-LIBERO with lerobot-eval --env.type=libero. | ev-vlajepa-paper-libero, ev-lerobot-vlajepa-libero |
| Teleop + record | no verified evidence | — |
Evidence 2 sources; the cell shows the highest tier
- T1 · upstream documentedpapersource 2026-02-10verified 2026-10-25 (today)
VLA-JEPA paper, Sec. 4.2: VLA-JEPA is evaluated on the LIBERO simulation benchmark, which the paper states uses the Franka Emika Panda arm.
“The LIBERO benchmark employs the Franka Emika Panda arm and comprises four task suites designed to facilitate research on lifelong learning in robotic manipulation.”
https://arxiv.org/abs/2602.10098 - T1 · upstream documentedofficial-docssource 2026-08-03verified 2026-10-25 (today)lerobot 0.6.1
LeRobot v0.6.1 VLA-JEPA doc: 'Reproducing the LIBERO results' section with a lerobot-train command on HuggingFaceVLA/libero and lerobot-eval of lerobot/VLA-JEPA-LIBERO on the LIBERO env.
“starts the training from the Pretrain checkpoint, trains for 30k steps on the LIBERO dataset.”
https://github.com/huggingface/lerobot/blob/v0.6.1/docs/source/vla_jepa.mdx
Embed this cell
Badge for READMEs and model cards. It always shows the current state.
[](https://caniembody.pages.dev/compat/franka-panda/vla-jepa)<iframe src="https://caniembody.pages.dev/embed/franka-panda/vla-jepa" width="380" height="132" style="border:0" title="VLA-JEPA on Franka Emika Panda | CanIEmbody"></iframe>History
- 2026-10-25 fecccd1 Initial release: robot policy compatibility matrix · Danilo Znamerovszkij