WAM-OPD: On-Policy Distillation for World Action Models
On the map: World models → Embodied and robot world models
Abstract (arXiv)
World Action Models (WAMs) generate both future video and robot actions, offering two connected outputs for post-training supervision. How can a pretrained WAM learn from a stronger Teacher on the histories it encounters during execution? We present WAM-OPD, which collects Student rollout histories and queries a Teacher for paired video and action targets. The Student learns from both targets while retaining its one-step video and action generation at deployment. Across 12 RoboTwin 2.0 tasks, WAM-OPD improves average success from 33.8% to 65.7%; across four real-robot tasks, it improves average success from 51.4% to 64.6%. With the collected Student histories held fixed, joint video-action supervision achieves the highest observed success on all three ablation tasks, while either modality alone also improves performance. A separate comparison with Teacher-generated histories finds task-dependent differences between the two history sources. These results demonstrate the value of paired video-action supervision for improving WAM policies without increasing their deployed sampling budget.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Embodied and robot world models, one of 11 directions in World models. Builds on 14 of its 43 references that are in the map. First posted on arXiv 23 Aug 2026.
Builds on
14 of 43 references.
Closest ancestors · 3
Same direction, newest first
- 2026WorldSample: Closed-loop Real-robot RL with World Modelling2 Jul
- 2026Flash-WAM: Modality-Aware Distillation for World Action Models3 Jun
- 2026Hi-WM: Human-in-the-World-Model for Scalable Robot Post-Training23 Apr
Other references · 11
- 2025RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulationcited by 477
- 2026Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillationcited by 71
- 2026DiffusionOPD: A Unified Perspective of On-Policy Distillation in Diffusion Modelscited by 35
- 2026Flow-OPD: On-Policy Distillation for Flow Matching Modelscited by 29
- 2025π_textttRL: Online RL Fine-tuning for Flow-based Vision-Language-Action Modelscited by 16
- 2026VLA-OPD: Bridging Offline SFT and Online RL for Vision-Language-Action Models via On-Policy Distillationcited by 9
- 2026Guided Action Flow: Q-Guided Inference for Flow-Matching Vision-Language-Action Policiescited by 4
- 2026WCM: A World Critic Model for Vision-Language-Action Reinforcement Learningcited by 4
- 2026Disturbance observer-based tracking control for roll-to-roll slot die coating systems under gap and pump rate disturbancescited by 3
- 2026SignVLA: A Gloss-Free Vision-Language-Action Framework for Real-Time Sign Language-Guided Robotic Manipulationcited by 3
- 2026WAM-RL: World-Action Model Reinforcement Learning with Reconstruction Rewards and Online Video SFTcited by 3
Built on by
Citing papers per year: 2026: 2.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2026WAM-OPD: Sharpening World Action Models via On-Policy Distillation
- 2026Vid2WAM: Distilling Video Diffusion Priors into World Action Models
- 2026EVO-WAM: Evolving World Action Models through Video-Action Verification
- 2026R2-WAM: Repair-and-Reject Post-Training for World Action Models
- 2026Enhancing Policy Learning with World-Action Model
- 2026GigaWorld-Policy-0.5: A Faster and Stronger WAM Empowered by AutoResearch
- 2026GigaWorld-Policy: An Efficient Action-Centered World--Action Model
- 2026OpenWAM: An Open Framework for Composable World-Action Models