DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhang, Wenyao, Liu, Hongsi, Qi, Zekun, Wang, Yunnan, Yu, Xinqiang, Zhang, Jiazhao, Dong, Runpei, He, Jiawei, Lu, Fan, Wang, He, Zhang, Zhizheng, Yi, Li, Zeng, Wenjun, Jin, Xin
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918130422382592
author Zhang, Wenyao
Liu, Hongsi
Qi, Zekun
Wang, Yunnan
Yu, Xinqiang
Zhang, Jiazhao
Dong, Runpei
He, Jiawei
Lu, Fan
Wang, He
Zhang, Zhizheng
Yi, Li
Zeng, Wenjun
Jin, Xin
author_facet Zhang, Wenyao
Liu, Hongsi
Qi, Zekun
Wang, Yunnan
Yu, Xinqiang
Zhang, Jiazhao
Dong, Runpei
He, Jiawei
Lu, Fan
Wang, He
Zhang, Zhizheng
Yi, Li
Zeng, Wenjun
Jin, Xin
contents Recent advances in vision-language-action (VLA) models have shown promise in integrating image generation with action prediction to improve generalization and reasoning in robot manipulation. However, existing methods are limited to challenging image-based forecasting, which suffers from redundant information and lacks comprehensive and critical world knowledge, including dynamic, spatial and semantic information. To address these limitations, we propose DreamVLA, a novel VLA framework that integrates comprehensive world knowledge forecasting to enable inverse dynamics modeling, thereby establishing a perception-prediction-action loop for manipulation tasks. Specifically, DreamVLA introduces a dynamic-region-guided world knowledge prediction, integrated with the spatial and semantic cues, which provide compact yet comprehensive representations for action planning. This design aligns with how humans interact with the world by first forming abstract multimodal reasoning chains before acting. To mitigate interference among the dynamic, spatial and semantic information during training, we adopt a block-wise structured attention mechanism that masks their mutual attention, preventing information leakage and keeping each representation clean and disentangled. Moreover, to model the conditional distribution over future actions, we employ a diffusion-based transformer that disentangles action representations from shared latent features. Extensive experiments on both real-world and simulation environments demonstrate that DreamVLA achieves 76.7% success rate on real robot tasks and 4.44 average length on the CALVIN ABC-D benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2507_04447
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge
Zhang, Wenyao
Liu, Hongsi
Qi, Zekun
Wang, Yunnan
Yu, Xinqiang
Zhang, Jiazhao
Dong, Runpei
He, Jiawei
Lu, Fan
Wang, He
Zhang, Zhizheng
Yi, Li
Zeng, Wenjun
Jin, Xin
Computer Vision and Pattern Recognition
Robotics
Recent advances in vision-language-action (VLA) models have shown promise in integrating image generation with action prediction to improve generalization and reasoning in robot manipulation. However, existing methods are limited to challenging image-based forecasting, which suffers from redundant information and lacks comprehensive and critical world knowledge, including dynamic, spatial and semantic information. To address these limitations, we propose DreamVLA, a novel VLA framework that integrates comprehensive world knowledge forecasting to enable inverse dynamics modeling, thereby establishing a perception-prediction-action loop for manipulation tasks. Specifically, DreamVLA introduces a dynamic-region-guided world knowledge prediction, integrated with the spatial and semantic cues, which provide compact yet comprehensive representations for action planning. This design aligns with how humans interact with the world by first forming abstract multimodal reasoning chains before acting. To mitigate interference among the dynamic, spatial and semantic information during training, we adopt a block-wise structured attention mechanism that masks their mutual attention, preventing information leakage and keeping each representation clean and disentangled. Moreover, to model the conditional distribution over future actions, we employ a diffusion-based transformer that disentangles action representations from shared latent features. Extensive experiments on both real-world and simulation environments demonstrate that DreamVLA achieves 76.7% success rate on real robot tasks and 4.44 average length on the CALVIN ABC-D benchmarks.
title DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2507.04447