Chain-of-Action: Trajectory Autoregressive Modeling for Robotic Manipulation
About
We present Chain-of-Action (CoA), a novel visuo-motor policy paradigm built upon Trajectory Autoregressive Modeling. Unlike conventional approaches that predict next step action(s) forward, CoA generates an entire trajectory by explicit backward reasoning with task-specific goals through an action-level Chain-of-Thought (CoT) process. This process is unified within a single autoregressive structure: (1) the first token corresponds to a stable keyframe action that encodes the task-specific goals; and (2) subsequent action tokens are generated autoregressively, conditioned on the initial keyframe and previously predicted actions. This backward action reasoning enforces a global-to-local structure, allowing each local action to be tightly constrained by the final goal. To further realize the action reasoning structure, CoA incorporates four complementary designs: continuous action token representation; dynamic stopping for variable-length trajectory generation; reverse temporal ensemble; and multi-token prediction to balance action chunk modeling with global structure. As a result, CoA gives strong spatial generalization capabilities while preserving the flexibility and simplicity of a visuo-motor policy. Empirically, we observe CoA achieves the state-of-the-art performance across 60 RLBench tasks and 8 real-world manipulation tasks.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Robot Manipulation | RLBench 10 | Success Rate97 | 20 | |
| Bimanual Robot Manipulation | RoboTwin 2.0 | Success Rate: Handover Block84 | 14 | |
| Robot Manipulation | Robomimic image observation | -- | 9 | |
| Pour Water | Real-robot tabletop tasks 15 rollouts | Success Rate33.3 | 4 | |
| Stack bowls | Real-robot tabletop tasks 15 rollouts | Success Rate26.7 | 4 | |
| tidy-up-desk | Real-robot tabletop tasks 15 rollouts | Success Rate (%)40 | 4 |