TAG: Target-Agnostic Guidance for Stable Object-Centric Inference in Vision-Language-Action Models
2026-03-25 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionRobotics
AI summaryⓘ
The authors studied how robots sometimes fail to pick the right object in messy scenes because they get confused by nearby items. They found that the robot’s hand often goes to the wrong place, even if the motion itself is possible. To fix this, they created a method called TAG that helps the robot focus better on the correct object by comparing what it sees with and without the object. This approach improves the robot’s accuracy without needing to change how the robot learns and works. Testing showed that TAG helps robots do better in cluttered environments and pick the right things more reliably.
Vision-Language-Action policiesinstance-level groundingclassifier-free guidancerobotic manipulationcluttered scenesaffordance groundinginference-time guidanceresidual steeringgrasp trajectoryrobotic benchmarks
Authors
Jiaying Zhou, Zhihao Zhan, Ruifeng Zhai, Qinhan Lyu, Hao Liu, Keze Wang, Liang Lin, Guangrun Wang
Abstract
Vision--Language--Action (VLA) policies have shown strong progress in mapping language instructions and visual observations to robotic actions, yet their reliability degrades in cluttered scenes with distractors. By analyzing failure cases, we find that many errors do not arise from infeasible motions, but from instance-level grounding failures: the policy often produces a plausible grasp trajectory that lands slightly off-target or even on the wrong object instance. To address this issue, we propose TAG (Target-Agnostic Guidance), a simple inference-time guidance mechanism that explicitly reduces distractor- and appearance-induced bias in VLA policies. Inspired by classifier-free guidance (CFG), TAG contrasts policy predictions under the original observation and an object-erased observation, and uses their difference as a residual steering signal that strengthens the influence of object evidence in the decision process. TAG does not require modifying the policy architecture and can be integrated with existing VLA policies with minimal training and inference changes. We evaluate TAG on standard manipulation benchmarks, including LIBERO, LIBERO-Plus, and VLABench, where it consistently improves robustness under clutter and reduces near-miss and wrong-object executions.