Vision language action policies improve task success with intent aware visual filtering
DIVA: Dual-Space Intent-Aware Visual Attenuation for Vision-Language-Action Policies
Artificial Intelligence
Summary
Vision-language-action systems help robots or AI perform tasks by understanding both images and language instructions. The authors created a method called DIVA that decides which parts of an image are important based on the task at hand, so the AI can focus on relevant details. This helps improve how well the AI completes tasks, even when the visual scene has distracting elements. The method works without needing extra labels about what things are in the scene and was tested in simulated and real-world settings.
What this means in practice
- •For robotics developers: Improve robot task completion by enabling policies to focus on visually relevant features aligned with the current intent.
- •For autonomous system engineers: Enhance robustness of vision-language controllers to distracting or irrelevant visual changes when executing intended actions.
Authors
Kaixi Feng, Guoheng Sun, Ziyao Wang, Yexiao He, Zheyu Shen, Ang Li
Abstract
Vision-language-action (VLA) policies typically feed dense visual patch tokens into a language-action backbone, preserving scene context but offering no explicit mechanism to regulate how strongly different visual tokens influence policy computation. We introduce DIVA, a Dual-Space Intent-Aware Visual Attenuation module with an anchor-then-attenuate design. DIVA combines high-level task intent with low-level visual evidence to estimate patch-wise relevance anchors, then applies them in two complementary spaces: it reweights projected visual tokens before backbone entry and persistently attenuates low-relevance visual states within the backbone. DIVA preserves the full visual token sequence and requires no external grounding supervision. On LIBERO, DIVA improves OpenVLA-OFT from 96.6% to 98.0% average success and raises its zero-shot LIBERO-Plus score from 69.6 to 72.6. Real-world experiments further show consistent gains under task-irrelevant visual perturbations, supporting the robustness of intent-aware visual attenuation beyond simulation.