LeWAM: Lightweight World Action Model Hits 92.28% Success on RoboTwin 2.0
udmrzn · x · 2026-09-25
Researchers introduced LeWAM (Latent evolving World Action Model), a lightweight world action model that unifies action generation and future visual prediction in I-JEPA's latent space without video diffusion. With AdaFuse and DemoDPO, it achieves a 92.28% average success rate on the RoboTwin 2.0 benchmark, showing latent-space prediction is a competitive alternative to video-diffusion-based world models for robot manipulation.
More from Embodied
- Disney Research shows how it trains robots to dance at monthly robotics AI paper club — DominiqueCAPaul · 2026-09-25
- NVIDIA pitches GPU-accelerated physics simulation as the training ground for real-world robots — LexiLove · 2026-09-25
- Community vote crowns Physical Intelligence as the robotics lab with top talent density — ZeYanjie · 2026-09-25
- AMD and Perplexity team up to bring Perplexity Portable Computer to Ryzen AI Halo — AnushElangovan · 2026-09-25
- Google Glass founder: glasses that block eye contact will never go mainstream — Scobleizer · 2026-09-25
- 19 Unitree Humanoids Dance With 120 Performers Before 10,000+ in Shanghai's Record Live Show — rohanpaul_ai · 2026-09-25