Event
Proposal of the AW-0 framework to unify code planning and video action prediction
2 reports · 2 sources
The study proposes the AW-0 agent world action model framework, which models code, future video, and robot actions as three coupled generative streams within a unified mixture-of-transformers architecture: a code expert generates parameterized skill routines and control logic, while video and action experts generate future visual states and low-level motor commands. The paper states that in evaluations on RoboTwin 2.0 and real-world dual-arm manipulation tasks, this framework achieved better task performance than VLA and WAM baselines; the paper also introduces an automatic curation pipeline that annotates aligned programs for robot and human video trajectories and consolidates recurring routines into a reusable skill library.
Report timeline
- AW-0: An Embodied World Action Model Framework Unifying Code-Based Planning and Video-Action Prediction
The paper proposes AW-0, an agentic world action model framework that combines code-based planning with joint video and action prediction. Within a unified mixture-of-transformers architecture, it simultaneously models three coupled generative streams: code, future video, and robot actions. A code expert generates parameterized skill routines and control logic, while video and action experts generate future visual states and low-level motor commands.
arXiv cs.RO ·
- AW-0: An Agentic World Action Model Framework Coupling Code Planning with Video Action Prediction
This work proposes AW-0, an agentic world action model framework that models code, future video, and robot actions as three coupled generative streams within a unified mixture-of-transformers architecture. A code expert generates parameterized skill routines and control logic, while video and action experts generate future visual states and low-level motor commands.
HuggingFace Daily Papers ·