ModAR: First Multimodal Autoregressive World-Action Model Cuts Training Compute 20x
Adam Hung · hf · 2026-09-16
ModAR is the first world-action model (WAM) to autoregressively denoise multiple future modalities before predicting actions, letting each prediction condition on previously generated modalities.
- Key finding: predicting point tracks, DINO features, and depth maps helps WAMs; additionally predicting future RGB offers no consistent benefit.
- Performance: ModAR's sequential generation outperforms existing WAM formulations with the highest average success rate at all data scales.
- Efficiency: fine-tuned on the video-initialized Flex-π, ModAR achieves a slightly higher success rate (75% vs 72%) with 20x fewer training FLOPs and no pretraining.
- Real world: outperforms baselines on three real bimanual tasks and improves with human videos.
More from Embodied
- Wearing Whoop and Apple Watch on a single dual-slot band — tinyfool · 2026-11-02
- REA: async brain-cerebellum runtime for robots where policies never wait on the LLM — Pengfei Zhou · 2026-11-02
- Robot Combat Sports Coming to Japan with a Permanent Arena in Tokyo — cixliv · 2026-11-02
- REK robot fighting in the cage: humanoid bot in boxing gloves draws a crowd — broodsugar · 2026-11-02
- iPhone Duo: a folding iPhone mod impresses, but its hinge raises crackgate concerns — rudrank · 2026-11-02
- InflatableBots: University of Calgary's inflatable robots bring dynamic haptics to large-scale VR — lukas_m_ziegler · 2026-11-02