AI summaryⓘ
The authors propose a method called Gated Reprojection for robot control models that helps decide when to spend extra computing time to improve robot actions. They use a technique (Best-of-N) that picks the best predicted future actions by checking if the robot’s predicted views are consistent from different angles, using a fixed geometry model. Their gating method only activates extra computations when the first guess looks inconsistent, saving time while still improving success rates across several robot tasks. They also find that checking reprojection consistency is a reliable way to select better actions without needing task-specific feedback. However, sometimes the method wrongly picks less ideal actions, explaining why more samples do not always help.
robot controlfoundation modelstest-time scalingBest-of-Nrolloutscross-view reprojectiongeometry foundation modelselective computationtask successaction-future consistency gate
Authors
Zesen Zhao, Minkyoung Cho, Hui shen, Boyuan Zheng, Kunxiao Gao, Yulong Cao, Z. Morley Mao
Abstract
Test-time scaling improves foundation-model inference by spending additional computation, but robot control requires deciding whether extra compute is useful before executing an action. World Action Models (WAMs) make this decision natural: each rollout exposes both an action chunk and predicted future observations. We propose \methodgated, a training-free selective test-time scaling framework for WAMs. We first instantiate \method, a fixed-budget Best-of-$N$ selector that ranks sampled rollouts by cross-view depth reprojection consistency of their predicted futures, computed with a frozen geometry foundation model. \methodgated\ adds a lightweight action--future consistency gate that invokes \method\ only when the initial rollout appears internally inconsistent. Across five benchmark--backbone settings on RoboCasa, LIBERO Long, and RoboTwin~2.0, fixed-budget \method\ improves $N{=}8$ task success in every setting, e.g., raising the RoboCasa group average from $66.3\%$ to $68.4\%$ with Cosmos Policy and from $80.8\%$ to $82.5\%$ with X-WAM. With gating enabled, \methodgated\ recovers on average $74.8\%$ of the always-on success gain while triggering additional sampling on only $26.2\%$ of decision points. Offline diagnostics show that cross-view reprojection is a strong task-label-free selector, and we identify false low-score selections as a failure mode that helps explain why performance can saturate or degrade as $N$ increases.