Berkeley and NVIDIA's AW-0 hits 80.8% on RoboTwin and 71.1 rubric score on real robots
trevordarrell · x · 2026-10-27
AW-0, a collaboration across UC Berkeley, NVIDIA, and UPenn, provides a shared foundation for planning, prediction, and execution: generated code serves both as a planning condition for learned control and an executable interface for skill reuse. It achieves 80.8% success on 50 RoboTwin 2.0 tasks (vs 64.8% for the strongest baseline) and a 71.1 average rubric score across 8 real-world tasks with YAM grippers and Tianji-Sharpa dexterous hands (vs 46.0 baseline). Paper, code, and models are open, pointing toward models that revise programs from execution feedback and expand skill libraries through physical interaction.
Related event: UC Berkeley and NVIDIA Unveil AW-0, a Unified World Action Model(4 posts)→
More from Embodied
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Bedrock Robotics raises $600M as its fully autonomous excavators do paid work on live job sites — Sethwinterroth · 2026-10-27
- AW-0 hits 80.8% on RoboTwin tasks, beating strongest baseline by 16 points — trevordarrell · 2026-10-27
- AW-0's code stream fixes world model prediction: picks mango where baseline grabs apple — trevordarrell · 2026-10-27
- AW-0 runs at 230ms on a single RTX 5090, 2.4x faster than autoregressive code planning — trevordarrell · 2026-10-27
- AW-0 debuts: one generative world action model plans in code, predicts, and controls robots — trevordarrell · 2026-10-27