Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911737245073408 |
|---|---|
| author | Liang, Zhixuan Li, Yizhuo Yang, Tianshuo Wu, Chengyue Mao, Sitong Pei, Liuao Nian, Tian Zhou, Shunbo Yang, Xiaokang Pang, Jiangmiao Mu, Yao Luo, Ping |
| author_facet | Liang, Zhixuan Li, Yizhuo Yang, Tianshuo Wu, Chengyue Mao, Sitong Pei, Liuao Nian, Tian Zhou, Shunbo Yang, Xiaokang Pang, Jiangmiao Mu, Yao Luo, Ping |
| contents | Vision-Language-Action (VLA) models adapt large vision-language backbones to map images and instructions into robot actions. However, prevailing VLAs either generate actions autoregressively in a fixed left-to-right order with poor performance or attach separate diffusion heads outside the backbone that fragments information pathways and hinders unified, scalable architectures. Instead, we present Discrete Diffusion VLA that discretizes action chunks and models them with discrete diffusion pattern retaining progressive refinement inside the unified transformer backbone. Our method achieves an adaptive decoding order that resolves high-confidence action elements before harder ones and employs secondary re-masking to revisit uncertain predictions, enabling robust error correction. This design preserves pretrained vision-language priors, supports parallel decoding, and improves the efficiency. Discrete Diffusion VLA achieves 96.4% avg. success on LIBERO, 71.2% visual matching on SimplerEnv-Fractal, and 54.2% overall on SimplerEnv-Bridge. On out-of-distribution tests of LIBERO-Goal, our method exhibits only 0.8% language degradation versus 8.0% of parallel decoding, and 20.4% vision degradation versus 29.0% for continuous diffusion, demonstrating well retention of pretrained vision-language capabilities. We also conduct two real-robot evaluations on AgileX Cobot Magic platform to show the method's effectiveness. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_20072 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies Liang, Zhixuan Li, Yizhuo Yang, Tianshuo Wu, Chengyue Mao, Sitong Pei, Liuao Nian, Tian Zhou, Shunbo Yang, Xiaokang Pang, Jiangmiao Mu, Yao Luo, Ping Computer Vision and Pattern Recognition Machine Learning Robotics Vision-Language-Action (VLA) models adapt large vision-language backbones to map images and instructions into robot actions. However, prevailing VLAs either generate actions autoregressively in a fixed left-to-right order with poor performance or attach separate diffusion heads outside the backbone that fragments information pathways and hinders unified, scalable architectures. Instead, we present Discrete Diffusion VLA that discretizes action chunks and models them with discrete diffusion pattern retaining progressive refinement inside the unified transformer backbone. Our method achieves an adaptive decoding order that resolves high-confidence action elements before harder ones and employs secondary re-masking to revisit uncertain predictions, enabling robust error correction. This design preserves pretrained vision-language priors, supports parallel decoding, and improves the efficiency. Discrete Diffusion VLA achieves 96.4% avg. success on LIBERO, 71.2% visual matching on SimplerEnv-Fractal, and 54.2% overall on SimplerEnv-Bridge. On out-of-distribution tests of LIBERO-Goal, our method exhibits only 0.8% language degradation versus 8.0% of parallel decoding, and 20.4% vision degradation versus 29.0% for continuous diffusion, demonstrating well retention of pretrained vision-language capabilities. We also conduct two real-robot evaluations on AgileX Cobot Magic platform to show the method's effectiveness. |
| title | Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies |
| topic | Computer Vision and Pattern Recognition Machine Learning Robotics |
| url | https://arxiv.org/abs/2508.20072 |