Driving · 2026-10-14
LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts
Wenxuan Guo, Yingping Liang, Zhaoguo Wang, Ziyuan Li, Angyuan Ma, Zhenyu Wu, Chengkun Wang, Bingyao Yu, Ze Chen, Borui Zhang, Xiuwei Xu, Meng Zhang, Erjin Zhou, Jiwen Lu
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
LoopWAM is a closed-loop world-action model that learns directly from mixed-quality robot deployment rollouts (successes, suboptimal behavior, and failures). It uses source-dependent noise sampling, quality steering, and quality-weighted action supervision while training video and action tokens together through parallel denoising. On RoboChallenge Table30-V2 it improves success from 23.67% to 43.33% across three deployment-training rounds.
Why it matters
The authors study how a generalist robot policy can improve from deployment rather than only from expert demonstrations. They address the problem that naïve action imitation can reinforce failures, while filtering failures discards interaction evidence. LoopWAM aims to keep full video supervision across trajectory qualities while regulating action imitation, and to allow learned consequences to directly participate in action generation.
Method
- Learns a joint world–action model with bidirectional attention and a shared parallel denoising process where video and action tokens interact throughout generation.
- Uses source-dependent noise sampling so recorded rollout actions become accessible when predicting their observed visual consequences, while keeping joint video–action generation in one stage.
- Applies quality steering (condition on trajectory quality) and quality-weighted action supervision (downweight action imitation for lower-quality trajectories, with failed trajectories receiving no action supervision).
Abstract (from arXiv)
Scaling expert demonstrations alone does not enable generalist robot policies to improve from deployment. Real-world rollouts contain successes, suboptimal behavior, and failures: imitating every recorded action can reinforce poor behavior, while filtering out failures discards evidence of physical interactions and their consequences. We introduce \textbf{LoopWAM}, a closed-loop world-action model that learns directly from this full quality spectrum, using rollouts from both itself and other deployed policies. Source-dependent noise sampling unifies behavior learning and action-conditioned world modeling, while quality steering and quality-weighted supervision retain full video supervision across qualities and regulate action imitation. Video and action tokens interact throughout parallel denoising, allowing learned consequences to inform action generation directly. On RoboChallenge Table30-V2, three deployment-and-training rounds raise success from 23.67\% to 43.33\%, a 19.7-percentage-point gain, with a final score of 56.80. LoopWAM also achieves 94.5\% average success on RoboTwin~2.0 and 49.5\% overall success on RoboCasa365 without additional embodied pretraining, obtaining the best overall results among the compared methods across all three benchmarks. On challenging shirt folding task, two rounds of deployment-driven improvement after single-task post-training raise success from 13.20\% to 55.56\%. These results demonstrate closed-loop improvement from mixed-quality experience under both multi-task learning and single-task post-training. \href{https://loop-wam.github.io/}{\textcolor{orange}{Project Page}}
Related papers
- $R^2$-WAM: Repair-and-Reject Post-Training for World Action Models
- RAP-WAM: Learning World-Action Models through Predictive Rehearsal and Action Perturbations
- EVO-WAM: Evolving World Action Models through Video-Action Verification
- OpenWAM: An Open Framework for Composable World-Action Models
- Right Ranking, Wrong Scale: Reproducing the Robotic World Model's Dynamics Model and Testing the Uncertainty of Its Follow-up
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC
LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts
Preprint posted: 2026-10-13 · Original record ↗
- arxiv
- 2610.18873
Explicit source versions (1)
- v2 · 2026-10-15
- Explicit source version detected · v2
Source date: 2026-10-15 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:id explicit version (official metadata)
- Preprint available
Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Scaling expert demonstrations alone does not enable generalist robot policies to improve from deployment. Real-world rollouts contain successes, suboptimal behavior, and failures: imitating every recorded action can reinforce poor behavior, while filtering out failures discards evidence of physical interactions and their consequences. We introduce LoopWAM, a closed-loop world-action model that learns directly from this full quality spectrum, using rollouts from both itself and other deployed policies. Source-dependent noise sampling unifies behavior learning and action-conditioned world modeling, while quality steering and quality-weighted supervision retain full video supervision across qualities and regulate action imitation. Video and action tokens interact throughout parallel denoising, allowing learned consequences to inform action generation directly. On RoboChallenge Table30-V2, three deployment-and-training rounds raise success from 23.67% to 43.33%, a 19.7-percentage-point gain, with a final score of 56.80. LoopWAM also achieves 94.5% average success on RoboTwin 2.0 and 49.5% overall success on RoboCasa365 without additional embodied pretraining, obtaining the best overall results among the compared methods across all three benchmarks. On challenging shirt folding task, two rounds of deployment-driven improvement after single-task post-training raise success from 13.20% to 55.56%. These results demonstrate closed-loop improvement from mixed-quality experience under both multi-task learning and single-task post-training. Project page: https://loop-wam.github.io/
