World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetry

2026-04-02

AI Summary

This work addresses the limited robustness of existing general-purpose world models when predicting under a large number of suboptimal actions. The authors propose decoupling action-conditioned state prediction into two factors: state plausibility and action reachability. A forward–inverse asymmetric verification mechanism enables self-correction and self-improvement by leveraging the abundance of action-free video data and the low-dimensional nature of action representations to enhance reliability in data-sparse regions. Furthermore, a diverse subgoal generator based on video corpora, combined with a sparse inverse dynamics model, enforces cycle consistency across subgoal–action–forward rollout sequences for validation. Evaluated across nine tasks in MiniGrid, RoboMimic, and ManiSkill, the approach achieves a 2× improvement in sample efficiency and an 18% gain in downstream policy performance.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Intelligent Robots: State Estimation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

General-purpose world models promise scalable policy evaluation, optimization, and planning, yet achieving the required level of robustness remains challenging. Unlike policy learning, which primarily focuses on optimal actions, a world model must be reliable over a much broader range of suboptimal actions, which are often insufficiently covered by action-labeled interaction data. To address this challenge, we propose World Action Verifier (WAV), a framework that enables world models to identify their own prediction errors and self-improve. The key idea is to decompose action-conditioned state prediction into two factors -- state plausibility and action reachability -- and verify each separately. We show that these verification problems can be substantially easier than predicting future states due to two underlying asymmetries: the broader availability of action-free data and the lower dimensionality of action-relevant features. Leveraging these asymmetries, we augment a world model with (i) a diverse subgoal generator obtained from video corpora and (ii) a sparse inverse model that infers actions from a subset of state features. By enforcing cycle consistency among generated subgoals, inferred actions, and forward rollouts, WAV provides an effective verification mechanism in under-explored regimes, where existing methods typically fail. Across nine tasks spanning MiniGrid, RoboMimic, and ManiSkill, our method achieves 2x higher sample efficiency while improving downstream policy performance by 18%.

Problem

• world models
• robustness
• suboptimal actions
• prediction reliability
• action-labeled data

Innovation

• world models
• self-improvement
• forward-inverse asymmetry
• cycle consistency
• action verification