NTU's ODE Policy hits 5.9x LIBERO success after one epoch, 10.8x faster inference
ZeYanjie · x · 2026-11-12
NTU's MARS Lab released ODE Policy, which learns action dynamics and integrates them over the robot's physical time instead of predicting raw action chunks — eliminating jumps at chunk boundaries by construction.
- 5.9x higher success than flow matching after one epoch (79.2%), 85% after two, while flow/diffusion baselines need 30+ epochs
- 10.8x faster inference, 5.4x zero-shot success under unseen perturbations, smoother trajectories without RTC
- Works as a visuomotor policy, as the action head of a VLA (π0.5), and in world action models; the Hugging Face team verified ODE + pi0.5/WAM fine-tuning on LIBERO variants
Paper, code and video are all public.
More from Embodied
- Hands-on RL-for-robotics workshop packs the room at electronica LIVE! event — ShawnHymel · 2026-11-12
- NeurIPS paper proves when RNN+MLP can provably recover true robot state — chriswolfvision · 2026-11-12
- PlotJuggler 4.2 now runs in your browser — facontidavide · 2026-11-12
- Tesla Robotaxi now runs 24/7 in every city where the service is available — Polymarket · 2026-11-12
- Lamport Robotics opens Pincer1 robot pre-orders: 100 units, $1,199, Solana payments accepted — R4nsack · 2026-11-12
- Uncertainty-Aware RWM: taming world-model hallucination for real robot control — breadli428 · 2026-11-12