Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models

TL;DR

This paper investigates the reliability of World Action Models by examining action-state consistency, proposing a new test-time selection strategy, and demonstrating improved success rates in robotic tasks.

Contribution

It introduces action-state consistency as a diagnostic for WAM reliability and proposes a value-free consensus method for better rollout selection.

Findings

Action-state consistency distinguishes successful and failed rollouts.

Background collapse can cause deceptive consistency in low-dynamics trajectories.

Consensus ranking improves success rates on RoboCasa and RoboTwin 2.0.

Abstract

World Action Models (WAMs) enable decision-making through imagined rollouts by predicting future observations and actions. However, the reliability of these imagined futures remains under-examined: is a generated future merely visually plausible, or is it dynamically compatible with the action sequence it claims to model? In this work, we identify action-state consistency, the alignment between predicted actions and induced state transitions, as a missing reliability axis for WAMs. Through a systematic study across representative joint-prediction and inverse-dynamics models, we find that action-state consistency systematically separates successful and failed rollouts across many tasks and follows similar success-failure trends as learned value estimates. These results suggest that consistency captures decision-relevant structure beyond visual realism. We further identify background collapse as an important boundary condition, where low-dynamics failed trajectories can become deceptively consistent because static futures are easier to predict. Building on these findings, we introduce a value-free consensus strategy for test-time selection, which ranks candidate rollouts by agreement among predicted futures. This strategy improves success rates on RoboCasa and RoboTwin 2.0 without additional training or reward modeling. Taken together, our findings establish action-state consistency as both a diagnostic tool for evaluating WAM reliability and a practical signal for value-free planning.