Contrastive Representation Regularization for Vision-Language-Action Models
This addresses the challenge of aligning vision-language representations with robotic signals for improved manipulation performance, representing an incremental advancement.
The paper tackled the problem of suboptimal representations in Vision-Language-Action models for robot manipulation by introducing Robot State-aware Contrastive Loss, which improved pick-and-place task success from 30.8% to 41.5% in simulation and from 45.0% to 58.3% on real-robot tasks.
Vision-Language-Action (VLA) models have shown its capabilities in robot manipulation by leveraging rich representations from pre-trained Vision-Language Models (VLMs). However, their representations arguably remain suboptimal, lacking sensitivity to robotic signals such as control actions and proprioceptive states. To address the issue, we introduce Robot State-aware Contrastive Loss (RS-CL), a simple and effective representation regularization for VLA models, designed to bridge the gap between VLM representations and robotic signals. In particular, RS-CL aligns the representations more closely with the robot's proprioceptive states, by using relative distances between the states as soft supervision. Complementing the original action prediction objective, RS-CL effectively enhances control-relevant representation learning, while being lightweight and fully compatible with standard VLA training pipeline. Our empirical results demonstrate that RS-CL substantially improves the manipulation performance of state-of-the-art VLA models; it pushes the prior art from 30.8% to 41.5% on pick-and-place tasks in RoboCasa-Kitchen, through more accurate positioning during grasping and placing, and boosts success rates from 45.0% to 58.3% on challenging real-robot manipulation tasks.