VANE: Reliable Test-Time Training for Vision-Language-Action Models via Future Visual Representation Prediction
2026-08-10 • Robotics
RoboticsComputer Vision and Pattern Recognition
AI summaryⓘ
The authors present a new method called VANE to improve how robots learn while they are working, especially when using vision, language, and action together. Their method carefully tests possible changes before applying them, only making updates when future outcomes confirm the change is helpful. This makes the robot's learning more reliable and reversible during tasks. They show improvements in success rates on simulated and real robot tests, although results vary by task and robot type.
Test-time trainingVision-language-action policiesRobot manipulationOnline adaptationSimulated environmentsRobot embodimentPolicy updatesClosed-loop control
Authors
Hongjin Ji, Guoyang Xia, Luoyang Sun, Fangxiang Feng, Lei Ren
Abstract
Test-time training (TTT) offers a lightweight way to adapt vision--language--action (VLA) policies from unlabeled deployment streams, but it remains difficult to use reliably in closed-loop manipulation. A shared adaptation space can mix incompatible task corrections, while an online update can alter subsequent actions before its consequences are known. We introduce a reliable TTT framework for VLA policies (VANE). VANE conditions prompt adaptation on the current vision--language context and learns from the future visual consequences of executed actions. Candidate updates are isolated from the live policy, evaluated on subsequent observations, and committed only when supported by future evidence, making adaptation selective and reversible. On SimplerEnv WidowX, VANE improves average success by $3.2$ percentage points over the corresponding TTT baseline. Results on Google Robot further show that deployment-time gains remain task- and embodiment-dependent. Together, these results demonstrate a constrained, evidence-based approach to adapting VLA policies during interaction.