A Causality-aware Infer-diagnose-refine Framework for Test-time Modality Adaptation in VLA Models
For robot manipulation tasks, the framework provides a training-free method to improve VLA model predictions by dynamically weighting visual observations at test time.
The paper tackles test-time modality adaptation in vision-language-action (VLA) models by proposing a causality-aware infer-diagnose-refine (IDR) framework that dynamically adjusts visual importance. Experiments show performance improvements across multiple VLA backbones on simulation benchmarks and real-world tasks.
Vision-language-action (VLA) models predict sequential actions to execute tasks specified by language instructions, conditioned on visual observations and proprioceptive states. However, how to fuse modalities in VLA models remains an open problem, since robot manipulation involves dynamic phases, such as long-distance movements and close-range interactions, in which the importance of visual observations may vary over time. In this paper, we propose an infer-diagnose-refine (IDR) framework, a model-agnostic framework that can be integrated with diverse VLA architectures for refining action predictions at test time. IDR first infers actions under factual and counterfactual scenarios of visual observations, and then diagnoses the causal effects of visual observations as the estimated dynamic importance, which is finally used to refine the action predictions in a training-free manner. We further design a causality-aware action refiner to realize the IDR framework, including zero-padding interventions for inferring counterfactual actions, norm-based quantification for diagnosing causal effects, and gated residual fusion for refining actions. Extensive experiments on both simulation benchmarks and real-world tasks show improvements in overall performance across multiple VLA backbones, demonstrating the efficacy of dynamically adjusting visual importance at test time.