# When Listening Becomes Easier: Scrubbing Visual Cues for Shortcut-Free VLAs

arXiv: https://arxiv.org/abs/2610.10912 · HTML: https://mlsanity.com/paper/2610.10912/

## Key facts

- No citing papers in the map yet (data through 7 Oct 2026).
- Builds on 17 of its 36 references that are in the map.
- First posted on arXiv 7 Oct 2026.

## Builds on (most cited first, 17 of 17)

- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Gaussian Error Linear Units (GELUs)](https://mlsanity.com/paper/1606.08415/) (2016)
- [SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics](https://mlsanity.com/paper/2506.01844/) (2025)
- [RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation](https://mlsanity.com/paper/2506.18088/) (2025)
- [π0.7: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities](https://mlsanity.com/paper/2604.15483/) (2026)
- [Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments](https://mlsanity.com/paper/2605.30280/) (2026)
- [JEPA-VLA: Video Predictive Embedding is Needed for VLA Models](https://mlsanity.com/paper/2602.11832/) (2026)
- [Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations](https://mlsanity.com/paper/2509.11417/) (2025)
- [Do You Need Proprioceptive States in Visuomotor Policies?](https://mlsanity.com/paper/2509.18644/) (2025)
- [Sparse Autoencoders Reveal Interpretable and Steerable Features in VLA Models](https://mlsanity.com/paper/2603.19183/) (2026)
- [VLANeXt: Recipes for Building Strong VLA Models](https://mlsanity.com/paper/2602.18532/) (2026)
- [LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models](https://mlsanity.com/paper/2603.28301/) (2026)
- [Breaking Lock-In: Preserving Steerability under Low-Data VLA Post-Training](https://mlsanity.com/paper/2604.23121/) (2026)
- [What Are We Actually Benchmarking in Robot Manipulation?](https://mlsanity.com/paper/2606.04233/) (2026)
- [Generalizable VLA Finetuning via Representation Anchoring and Language-Action Alignment](https://mlsanity.com/paper/2607.13429/) (2026)
- [AC-VLA: Robust Out-of-Distribution Action Execution via Compositional Learning](https://mlsanity.com/paper/2607.15714/) (2026)
- [Unmasking the Illusion of Embodied Reasoning in Vision-Language-Action Models](https://mlsanity.com/paper/2604.18000/) (2026)

## Abstract (arXiv)

Shortcut learning is a prevalent issue in robot learning. The limited diversity of robot demonstration datasets can mislead policies into exploiting spurious correlations between tasks and irrelevant features, such as viewpoint or background. Collecting sufficiently diverse robot demonstrations is costly and inefficient, motivating algorithmic alternatives. We focus on vision-language-action (VLA) models and discover that different vision-language model backbones exhibit substantially different levels of susceptibility to visual shortcut learning. We find that model behavior correlates with our proposed representation-level metric, action margin, which requires no policy rollouts. Visual shortcuts consistently enter action representations in early layers, with models differing in the extent to which later layers correct them by incorporating language information. To boost models' attention to language, we introduce task scrubbing, a new domain-adversarial training method that decreases models' likelihood of using visual shortcuts and improves VLAs' generalization. Experiments in both simulation and the real world across multiple VLAs and visual cues show that task scrubbing improves out-of-distribution robustness and often eliminates visual shortcut learning.