Vela: spline-based continuous trajectories scale vision-language-action models
Yifan Li · hf · 2026-10-15
Researchers from Fudan, CUHK, and NovaXBot introduce Vela, a vision-language-action foundation model that represents future robot behavior as continuous trajectories rather than fixed-rate action chunks.
- Pointwise action chunks waste capacity on correlated neighboring actions and force a tradeoff between long-horizon coverage and the local precision needed for contact-rich manipulation
- Vela combines a compact spline-based action representation, motion-dependent temporal support, and a shared action interface for heterogeneous embodiments
- Pretrained on large-scale multi-embodiment robot data, it shows promising results on LIBERO-X, EBench, and two real-world long-horizon tasks: egg-cake cooking and potato shredding
The results suggest continuous action representations as a foundation for future embodied foundation models.
More from Embodied
- Tesla adds 30 more Cybercabs in Texas, purpose-built fleet hits 349 with 300+ serving Austin — XFreeze · 2026-10-15
- Physical AI webinar Oct 22: four robotics founders on evals, retraining, and field failures — carrycooldude · 2026-10-15
- Scoble: healthcare robots arrive next year, true humanoids likely 2-7 years away — realmeetjames · 2026-10-15
- KAIST's BeetleBot: finger-controlled soft robot walks, grips and streams sensory feedback — CurieuxExplorer · 2026-10-15
- 14-year-old builds handheld AI pesticide scanner PestiSCAND with 85% accuracy — nordicinst · 2026-10-15
- Beat Proof of Quality, win a Tesla: robotics challenge offers five teams a shot — EnigmaFund · 2026-10-15