Robotics · 2026-10-05
Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation
Shukai Gong, Xuanran Zhai, Yintianrun Zhang, Ruopeng Cui, Ye Huang, Yiyang Fu, Dexuan Lyu, Chaojie Li, Xinyi Song, Peiwen Lin, Chuang Wang, Mingyuan Jia, Yufan Deng, Jiaxin Fang, Bo Liang, Jiaxin Li, Yuxiang Gao, Hao Liu, Daquan Zhou
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
ViGAR is a hierarchical world-action model for long-horizon compositional robotic manipulation. It factorizes task progress into a visual subgoal planner (predicts the next subtask goal image) and a subgoal executor (jointly predicts future visual trajectories and actions conditioned on the predicted subgoal). It supports in-context robotic manipulation by using a global goal image as inference-time context, without parameter updates.
Why it matters
The authors argue that existing world-action models provide local control but lack explicit task-level decision making for compositional subtasks. ViGAR introduces visual subgoal planning as an explicit decision variable and conditions world-action generation on it, aiming to improve compositional success and enable inference-time task composition changes via global goal images.
Method
- Subgoal planner: predicts a visual subgoal for the next subtask from the current observation and global instruction; the planner is trained as an image-editing problem with a flow-matching objective and end-effector ROI weighting.
- Subgoal-guided world-action model: conditioned on the predicted subgoal, jointly predicts future visual frames and an action chunk via a joint flow-matching objective over visual and action targets.
- In-context robotic manipulation: at inference, a user-specified global goal image is provided as context to the planner, inducing a different subtask decomposition without parameter updates.
Abstract (from arXiv)
Long-horizon compositional manipulation has become increasingly important for real-world robot deployment, where a single task involves multiple coordinated subtasks. Existing world-action models (WAMs) jointly predict short-horizon visual futures and actions, but typically lack explicit subtask-level reasoning. We propose Visual Goal-conditioned Action Reasoning (ViGAR), a hierarchical framework that factorizes manipulation into a visual subgoal planner and a subgoal executor. Given the current observation and global instruction, the subgoal planner predicts a visual subgoal for the next subtask. The subgoal executor then jointly generates future visual trajectories and actions conditioned on the predicted subgoal. Both components share a pretrained world-model representation, enabling task-level planning and action generation to benefit from common physical knowledge. Moreover, our framework naturally supports in-context learning: using a global goal image as context can induce different subtask decompositions and behaviors without parameter updates. On the RoboTwin Clean2Random benchmark, ViGAR achieves 82.00% and 67.02% success rates under the Clean and Random settings, respectively, surpassing the strongest baseline by 12.86 percentage points in average success rate. Real-world robot experiments on five compositional and two in-context learning tasks further confirm the effectiveness of ViGAR.
Related papers
- PILOT: Guiding Long-Horizon World-Model Planning for Robot Manipulation with Visual Subgoals and Policy Priors
- WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
- Agentic World Action Models
- Achieve What You Imagined: Learning to Align Actions with Visual Plans
- PINT: A Unified Model with Shared Predictive Intent for Robot Manipulation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-01 · Indexed by Pub-AI: 2026-10-05 UTC
Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation
Preprint posted: 2026-10-01 · Original record ↗
- arxiv
- 2610.02368
Explicit source versions (1)
- v1 · 2026-10-01
- Preprint available
Source date: 2026-10-01 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Long-horizon compositional manipulation has become increasingly important for real-world robot deployment, where a single task involves multiple coordinated subtasks. Existing world-action models (WAMs) jointly predict short-horizon visual futures and actions, but typically lack explicit subtask-level reasoning. We propose Visual Goal-conditioned Action Reasoning (ViGAR), a hierarchical framework that factorizes manipulation into a visual subgoal planner and a subgoal executor. Given the current observation and global instruction, the subgoal planner predicts a visual subgoal for the next subtask. The subgoal executor then jointly generates future visual trajectories and actions conditioned on the predicted subgoal. Both components share a pretrained world-model representation, enabling task-level planning and action generation to benefit from common physical knowledge. Moreover, our framework naturally supports in-context learning: using a global goal image as context can induce different subtask decompositions and behaviors without parameter updates. On the RoboTwin Clean2Random benchmark, ViGAR achieves 82.00% and 67.02% success rates under the Clean and Random settings, respectively, surpassing the strongest baseline by 12.86 percentage points in average success rate. Real-world robot experiments on five compositional and two in-context learning tasks further confirm the effectiveness of ViGAR.