What Matters, When? Diagnosing and Improving Conditional Visual Grounding in Visuomotor Imitation Policies

What Matters, When? Diagnosing and Improving Conditional Visual Grounding in Visuomotor Imitation Policies

|Authors: Vivek Chavan, Pengtao Xie, Yahuan Shi, Oliver Heimann, Kevin Haninger, Jörg Krüger

|Venue: arXiv preprint | Year: 2026

|arXiv: 2609.05376v1

Abstract

Visuomotor imitation policies can achieve high performance under in-distribution visual conditions yet fail when visually similar objects or receptacles are introduced. We study this behavior as a problem of conditional visual grounding: the visual target required for successful control changes with the manipulation phase and, in more complex tasks, with the observed task state. Using Action Chunking with Transformers (ACT), we systematically introduce distractor objects and receptacles with controlled color and shape similarity and localize failures to picking and placement. We find that distractor sensitivity is specific to both the type of visual similarity and the manipulation stage. Guided by this diagnosis, we evaluate distractor augmentation, phase-dependent attention regularization, and appearance-based visual prompting as complementary interventions for improving target selection while preserving spatial information required for control. These interventions substantially improve robustness in simulation and on a physical UR3e. We further examine the same failure pattern in a pretrained vision-language-action policy on a state-conditioned instrument-handling task, where the observed state of a medical instrument determines the correct destination. Together, the results show that visual distractors can cause incorrect object or destination selection even when the underlying manipulation skill remains intact, and that explicitly improving target selection can substantially recover performance across distinct visuomotor policy-learning regimes.

Key Contributions

• Visuomotor imitation policies can achieve high performance under in-distribution visual condition…
• We study this behavior as a problem of conditional visual grounding: the visual target required f…
• Using Action Chunking with Transformers (ACT), we systematically introduce distractor objects and…
• We find that distractor sensitivity is specific to both the type of visual similarity and the man…
• Guided by this diagnosis, we evaluate distractor augmentation, phase-dependent attention regulari…

Topics

• manipulation
• vision
• vla
• reinforcement-learning
• control

Code & Data

No code repository linked in paper metadata.

BibTeX

@article{Chavan2026_260905376v1,
  title     = {What Matters, When? Diagnosing and Improving Conditional Visual Grounding in Visuomotor Imitation Policies},
  author    = {Vivek Chavan and Pengtao Xie and Yahuan Shi and Oliver Heimann and Kevin Haninger and Jörg Krüger},
  year      = {2026},
  eprint    = {2609.05376v1},
  archivePrefix = {arXiv},
  primaryClass  = {cs.RO},
  url       = {https://arxiv.org/abs/2609.05376v1}
}