Search papers, labs, and topics across Lattice.
This paper introduces World-Coherent Decoding (WCD), a self-verifying framework for test-time planning that enhances the reliability of World Action Models (WAMs) in robotic control. By sampling multiple future-action hypotheses and ranking them based on visual plausibility and action stability, WCD effectively mitigates the impact of imagination-reality mismatches. The method achieved a significant performance improvement on the RoboTwin 2.0 benchmark, increasing success rates under limited supervision and demonstrating robustness in real-world scenarios.
Test-time planning for robots can be dramatically improved by focusing on selecting reliable future-action hypotheses rather than merely generating more of them.
World Action Models (WAMs) aim to control robots by stochastically generating visual futures and then decoding actions, but empirical observations indicate that the results can strongly depend on which future is selected. We propose World-Coherent-Decoding (WCD), a self-verifying test-time planning framework that treats WAM rollouts as falsifiable future--action hypotheses. At each decision step, WCD samples multiple candidates from a frozen WAM and ranks them using internal generative signals: flow-based video surprisal for visual plausibility and action path effort for action-generation stability. After execution, the realized observation audits the selected imagination, yielding an imagination--reality mismatch that trains a lightweight online predictor for future candidate selection. Thus, WCD converts delayed self-verification into pre-execution reliability estimation without updating the backbone model. On RoboTwin 2.0, WCD improves Hard success under limited randomized-scene supervision from $55.80\%$ to $60.90\%$, with a $+16.43$ gains on Horizon-3 tasks, and shows qualitative robustness on real Franka visual-shift tests. These results highlight a simple principle: test-time scaling for WAMs depends less on sampling more futures than on selecting reliable ones.
