Test-Time Adaptation of VLAs

VLA: Vision-Language-Action Model

Overview

This field studies how vision-language-action (VLA) robot policies can adapt during deployment to unfamiliar scenes, camera views, tasks, and robot execution errors. Approaches include test-time training from observed transitions, retrieval of demonstrations and memories, online reinforcement learning, and training-free steering or correction of frozen policies; uncertainty estimates can also allocate extra reasoning or control effort. Methods are commonly evaluated on manipulation benchmarks such as LIBERO, SIMPLER, and RoboTwin, alongside real-robot tests. A central tension is gaining reliable adaptation from limited or noisy feedback without excessive computation, degrading performance on familiar tasks, or making unsafe updates.

Latest papers

61 papers7 papers in the last four weeks, down 68% on the four weeks before. 0.1% of all new papers.

Feb 24, 2026

VLA Knows Its Limits: Adaptive Execution Horizons for Robot Policies

Action chunking has recently emerged as a standard practice in flow-based Vision-Language-Action (VLA) models. However, the effect and choice of the execution horizon - the number of actions to be executed from each predicted chunk - remains underexplored. In this work, we first show that varying the execution horizon leads to substantial performance deviations, with performance initially improving and then declining as the horizon increases. To uncover the reasons, we analyze the cross- and self-attention weights in flow-based VLAs and reveal two key phenomena: (i) intra-chunk actions attend invariantly to vision-language tokens, limiting adaptability to environmental changes; and (ii) the initial and terminal action tokens serve as stable anchors, forming latent centers around which intermediate actions are organized. Motivated by these insights, we interpret action self-attention weights as a proxy for the model's predictive limit and propose AutoHorizon, the first test-time method that dynamically estimates the execution horizon for each predicted action chunk to adapt to changing perceptual conditions. Across simulated and real-world robotic manipulation tasks, AutoHorizon is performant, incurs negligible computational overhead, and generalizes across diverse tasks and flow-based models.