https://tuagoale.github.io/tags/vision-language-action-models/Vision-Language-Action ModelsHugoBlox Kit (https://hugoblox.com)en-usSun, 26 Apr 2026 00:00:00 +0000https://tuagoale.github.io/media/icon.svghttps://tuagoale.github.io/tags/vision-language-action-models/https://tuagoale.github.io/publications/generative-control/Sun, 26 Apr 2026 00:00:00 +0000https://tuagoale.github.io/publications/generative-control/<p>GeCO addresses a structural inefficiency in diffusion and flow-matching policies: a fixed inference schedule spends the same computation on easy and difficult states. The method treats action synthesis as optimization over a learned stationary velocity field, allowing adaptive computation at test time.</p> <p>In addition to improving the success-latency trade-off, the field norm after optimization serves as an intrinsic OOD indicator. This makes the approach useful not only as a replacement for flow-matching action heads in VLA policies, but also as a mechanism for safer robotic deployment.</p> <p>Zunzhe Zhang and Runhan Huang contributed equally to this work.</p>