VIDEAS: Distilling Explicit Action Semantics from Demonstration Videos for World Models via Prior-Guided Simulation

2026-09-27

AI Summary

This study addresses the limitations of existing world models in internalizing causal physical laws and explicitly reasoning about action preconditions and state transitions by proposing a language-based world model construction framework. Methodologically, explicit action semantics are distilled from demonstration videos to endow the model with physical priors. A prior-guided trajectory simulation mechanism, combined with a negative sample augmentation strategy, is designed to effectively enhance data diversity and knowledge verification capabilities. Technically, the framework integrates vision-language models, textual environment simulation, and data distillation paradigms. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on embodied action semantic reasoning tasks, exhibiting profound physical understanding and strong generalization capabilities.

Abstract

World models learn internal representations of environment dynamics to predict future states, enabling agents to optimize action plans without physical interactions. However, developing world models that genuinely internalize underlying causal physical laws to explicitly reason about action preconditions and subsequent state transitions remains an open challenge. In this paper, we propose VIDEAS, a data distillation framework that transforms continuous physical dynamics from operational videos into explicit action semantics for foundation models. Specifically, it deconstructs visual demonstrations into discrete action trajectories and utilizes advanced vision-language models (VLMs) to extract structured knowledge encapsulating action preconditions and effects. To ensure physical consistency, we introduce a prior-guided trajectory simulation mechanism grounded within a text-based environment to rigorously validate the extracted knowledge. Notably, we incorporate negative trajectories to enrich knowledge completeness and enhance data diversity to mitigate cognitive bias. Furthermore, we present VIDEAS-WM, an 8B/9B-parameter suite of language-based world models trained on 34K high-quality samples derived from AgiBot-World dataset. Extensive experiments demonstrate that VIDEAS-WM establishes state-of-the-art performance in high-level embodied action semantic reasoning, exhibiting profound physical understanding and robust generalization across unseen scenarios.

Problem

• world models
• action semantics
• causal physical laws
• state transitions
• embodied reasoning

Innovation

• World Models
• Data Distillation
• Action Semantics
• Vision-Language Models
• Prior-Guided Simulation