PAVXploreRL is a reinforcement learning framework built on a pretrained latent world model that explicitly optimizes PAV objectives through reward-driven training, and jointly leverages ID trajectories and noise-driven OOD action exploration, without paired video supervision.
Abstract
Action-conditioned world models are a key component of embodied AI, serving as scalable policy evaluators that reduce reliance on expensive real-world rollouts. To accurately capture diverse action-induced dynamics, such models should satisfy three key objectives-Physical Plausibility (P), Action Adherence (A), and Visual Fidelity (V), collectively referred to as PAV-while remaining robust to both in-distribution (ID) expert demonstrations and out-of-distribution (OOD) actions. However, existing methods primarily rely on ID action-video pairs and pixel-level reconstruction losses, which do not explicitly optimize PAV objectives and generalize poorly beyond expert data. To address this, we propose PAVXploreRL, a reinforcement learning framework built on a pretrained latent world model that explicitly optimizes PAV objectives through reward-driven training. To improve action generalization, our method jointly leverages ID trajectories and noise-driven OOD action exploration, without paired video supervision. Experiments show that PAVXploreRL consistently outperforms pretrained baselines, achieving a 5.6% average gain across benchmarks and producing higher-quality PAV properties. As a policy evaluator, it also yields more reliable performance estimates and reduces the overestimation bias of prior expert-only world models such as Ctrl-World. Code: https://github.com/Social-AI-Studio/PAVXploreRL
Vid2WAM is proposed, an offline distillation framework that transfers visual diffusion priors from a large video foundation model into a compact WAM student and introduces source-aware residual action adaptation that learns source-specific corrections around a shared action backbone and mitigates interference from noisy pseudo-actions.
Chen-Hao Qiu, Ruixiang Wang, Runyi Zhao et al.· 0 citations
SimWAM, a simple yet effective WAM that leverages future-video prediction as a training-time supervision signal, co-trains a pretrained video expert and a lightweight action expert with joint flow matching and applies reinforcement learning to optimize a compositional driving reward beyond trajectory imitation.
LD4WAM is presented, which pairs a Latent Dynamics Model trained with semantic reconstruction and real motion alignment with a World Dynamics Action Model built as a mixture-of-transformers (MoT), which preserves full future-video generation and uses learnable queries to distill these latent dynamics from generated futures for action conditioning.
Zhen Shen, Jiaqi Liang, Jasper Lu et al.· 1 citation
Causal Action Effect Reweighting (CAER), a general training paradigm that redistributes supervision toward the tokens whose predicted future is causally affected by the action, is introduced.
Jianjie Fang, Xvyuan Liu, Zi-You Wang et al.· 0 citations
WAM-OPD, a deployment-consistent post-training recipe for a video-first WAM, is introduced and it is suggested that dense teacher supervision on student-induced histories is a promising post-training interface for video-first WAMs.
Liu Yang, Zhuang Jiang, Chenchao Sheng et al.· 1 citation
World-action models (WAM) predict future states to guide robot actions, enabling learning from both action-free video and action-labeled interaction. Most inherit pretrained video generators, leaving WAM pretraining and scaling underexplored. We introduce Genie Envisioner Act 2.0 (GE-Act 2.0), a world-action model whose trainable generative and action components are all initialized from scratch on manipulation data. It combines a control-oriented autoencoder (CoAE), a single-step visual planner (SVP), and an inverse dynamics model (IDM). CoAE retains action- and instruction-relevant information under aggressive compression, while SVP produces a complete future state in one differentiable pass, so visual planning and inverse dynamics can be pretrained separately on complementary data. The components are then jointly trained with knowledge-aligned selective optimization (KASO), which reduces mismatched supervision by selecting only predicted futures judged behaviorally compatible with the recorded action. We evaluate pretrained checkpoints directly, without per-task fine-tuning, on 100 tasks across 20 manipulation skill groups with held-out scenes, backgrounds, lighting, and object instances. Scaling co-training data from 300 to 30,000 hours raises success from 17.1% to 44.1% on G1-OP and from 13.4% to 31.1% on G2-90D; despite comprising less than 2% of the co-training data, G2-90D improves by 17.7 points, suggesting cross-embodiment transfer. Gains span 19/20 and 18/20 skill groups, and skill-specific coverage strongly correlates with zero-shot out-of-distribution (OOD) success (Pearson r=0.80; Spearman rho=0.85). Under the same protocol, the model grounds object, color, shape, and position references in at least 90% of trials and follows explicit instructions even when they conflict with an already-committed behavior or a conventional scene association.
AgiBot Research Team, Renhang Liu, Wen-Zhi Zhao et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
