Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhai, Andy, Liu, Brae, Fang, Bruno, Cai, Chalse, Ma, Ellie, Yin, Ethan, Wang, Hao, Zhou, Hugo, Wang, James, Shi, Lights, Liang, Lucy, Wang, Make, Wang, Qian, Gan, Roy, Yu, Ryan, Li, Shalfun, Liu, Starrick, Chen, Sylas, Chen, Vincent, Xu, Zach
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2509.11766
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909788604989440
author Zhai, Andy
Liu, Brae
Fang, Bruno
Cai, Chalse
Ma, Ellie
Yin, Ethan
Wang, Hao
Zhou, Hugo
Wang, James
Shi, Lights
Liang, Lucy
Wang, Make
Wang, Qian
Gan, Roy
Yu, Ryan
Li, Shalfun
Liu, Starrick
Chen, Sylas
Chen, Vincent
Xu, Zach
author_facet Zhai, Andy
Liu, Brae
Fang, Bruno
Cai, Chalse
Ma, Ellie
Yin, Ethan
Wang, Hao
Zhou, Hugo
Wang, James
Shi, Lights
Liang, Lucy
Wang, Make
Wang, Qian
Gan, Roy
Yu, Ryan
Li, Shalfun
Liu, Starrick
Chen, Sylas
Chen, Vincent
Xu, Zach
contents While foundation models show remarkable progress in language and vision, existing vision-language models (VLMs) still have limited spatial and embodiment understanding. Transferring VLMs to embodied domains reveals fundamental mismatches between modalities, pretraining distributions, and training objectives, leaving action comprehension and generation as a central bottleneck on the path to AGI. We introduce WALL-OSS, an end-to-end embodied foundation model that leverages large-scale multimodal pretraining to achieve (1) embodiment-aware vision-language understanding, (2) strong language-action association, and (3) robust manipulation capability. Our approach employs a tightly coupled architecture and multi-strategies training curriculum that enables Unified Cross-Level CoT-seamlessly unifying instruction reasoning, subgoal decomposition, and fine-grained action synthesis within a single differentiable framework. Our results show that WALL-OSS attains high success on complex long-horizon manipulations, demonstrates strong instruction-following capabilities, complex understanding and reasoning, and outperforms strong baselines, thereby providing a reliable and scalable path from VLMs to embodied foundation models.
format Preprint
id arxiv_https___arxiv_org_abs_2509_11766
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Igniting VLMs toward the Embodied Space
Zhai, Andy
Liu, Brae
Fang, Bruno
Cai, Chalse
Ma, Ellie
Yin, Ethan
Wang, Hao
Zhou, Hugo
Wang, James
Shi, Lights
Liang, Lucy
Wang, Make
Wang, Qian
Gan, Roy
Yu, Ryan
Li, Shalfun
Liu, Starrick
Chen, Sylas
Chen, Vincent
Xu, Zach
Robotics
While foundation models show remarkable progress in language and vision, existing vision-language models (VLMs) still have limited spatial and embodiment understanding. Transferring VLMs to embodied domains reveals fundamental mismatches between modalities, pretraining distributions, and training objectives, leaving action comprehension and generation as a central bottleneck on the path to AGI. We introduce WALL-OSS, an end-to-end embodied foundation model that leverages large-scale multimodal pretraining to achieve (1) embodiment-aware vision-language understanding, (2) strong language-action association, and (3) robust manipulation capability. Our approach employs a tightly coupled architecture and multi-strategies training curriculum that enables Unified Cross-Level CoT-seamlessly unifying instruction reasoning, subgoal decomposition, and fine-grained action synthesis within a single differentiable framework. Our results show that WALL-OSS attains high success on complex long-horizon manipulations, demonstrates strong instruction-following capabilities, complex understanding and reasoning, and outperforms strong baselines, thereby providing a reliable and scalable path from VLMs to embodied foundation models.
title Igniting VLMs toward the Embodied Space
topic Robotics
url https://arxiv.org/abs/2509.11766