Latent Reasoning VLA: Latent Thinking and Prediction for Vision-Language-Action Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914541587136512 |
|---|---|
| author | Bai, Shuanghao Lyu, Jing Zhou, Wanqi Li, Zhe Wang, Dakai Xing, Lei Zhao, Xiaoguang Wang, Pengwei Wang, Zhongyuan Chi, Cheng Chen, Badong Zhang, Shanghang |
| author_facet | Bai, Shuanghao Lyu, Jing Zhou, Wanqi Li, Zhe Wang, Dakai Xing, Lei Zhao, Xiaoguang Wang, Pengwei Wang, Zhongyuan Chi, Cheng Chen, Badong Zhang, Shanghang |
| contents | Vision-Language-Action (VLA) models benefit from chain-of-thought (CoT) reasoning, but existing approaches incur high inference overhead and rely on discrete reasoning representations that mismatch continuous perception and control. We propose Latent Reasoning VLA (LaRA-VLA), a unified VLA framework that internalizes multi-modal CoT reasoning into continuous latent representations for embodied action. LaRA-VLA performs unified reasoning and prediction in latent space, eliminating explicit CoT generation at inference time and enabling efficient, action-oriented control. To realize latent embodied reasoning, we introduce a curriculum-based training paradigm that progressively transitions from explicit textual and visual CoT supervision to latent reasoning, and finally adapts latent reasoning dynamics to condition action generation. We construct two structured CoT datasets and evaluate LaRA-VLA on both simulation benchmarks and long-horizon real-robot manipulation tasks. Experimental results show that LaRA-VLA consistently outperforms state-of-the-art VLA methods while reducing inference latency by up to 90\% compared to explicit CoT-based approaches, demonstrating latent reasoning as an effective and efficient paradigm for real-time embodied control. Project Page: https://loveju1y.github.io/Latent-Reasoning-VLA/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01166 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Latent Reasoning VLA: Latent Thinking and Prediction for Vision-Language-Action Models Bai, Shuanghao Lyu, Jing Zhou, Wanqi Li, Zhe Wang, Dakai Xing, Lei Zhao, Xiaoguang Wang, Pengwei Wang, Zhongyuan Chi, Cheng Chen, Badong Zhang, Shanghang Robotics Vision-Language-Action (VLA) models benefit from chain-of-thought (CoT) reasoning, but existing approaches incur high inference overhead and rely on discrete reasoning representations that mismatch continuous perception and control. We propose Latent Reasoning VLA (LaRA-VLA), a unified VLA framework that internalizes multi-modal CoT reasoning into continuous latent representations for embodied action. LaRA-VLA performs unified reasoning and prediction in latent space, eliminating explicit CoT generation at inference time and enabling efficient, action-oriented control. To realize latent embodied reasoning, we introduce a curriculum-based training paradigm that progressively transitions from explicit textual and visual CoT supervision to latent reasoning, and finally adapts latent reasoning dynamics to condition action generation. We construct two structured CoT datasets and evaluate LaRA-VLA on both simulation benchmarks and long-horizon real-robot manipulation tasks. Experimental results show that LaRA-VLA consistently outperforms state-of-the-art VLA methods while reducing inference latency by up to 90\% compared to explicit CoT-based approaches, demonstrating latent reasoning as an effective and efficient paradigm for real-time embodied control. Project Page: https://loveju1y.github.io/Latent-Reasoning-VLA/ |
| title | Latent Reasoning VLA: Latent Thinking and Prediction for Vision-Language-Action Models |
| topic | Robotics |
| url | https://arxiv.org/abs/2602.01166 |