20 results for “Pre-trained Vision-Language-Action models”
CS papers onlyHybrid search: Keyword + semantic, ranked by combined score.ⓘ
Want pure semantic search? Try claim verification →
Kechun Xu, Zhenjie Zhu, Anzhe Chen, Rong Xiong +1 more
This paper proposes a two-stage training method called APT to improve the generalization of Vision-Language-Action models to out-of-distribution language instructions.
Dong Jing, Tianqi Zhang, Jiaqi Liu, Jinman Zhao +4 more
This paper proposes a two-stage training framework to pretrain action modules with motion priors before Vision-Language-Action (VLA) alignment, improving VLA performance and reducing optimization chal…
The paper proposes Continuous Reasoning for Vision-Language-Action (VLA) models, arguing that effective reasoning must be a shared, verifiable internal latent space rather than discrete text tokens, l…
Rahul Sajnani, Diego Romeres, Mingyuan Zhang, Kai Gao +8 more
This paper introduces UniWALM, a unified model that integrates video, action, and language in a single representation, improving manipulation and memory-dependent behavior.
Pankhuri Vanjani, Zhuoyue Li, Jakub Suliga, Moritz Reuss +3 more
This paper presents DAM-VLA, a vision-language-action model that decouples temporal processing per modality to improve performance in contact-rich real-world manipulation tasks.
Junke Wang, Qihang Zhang, Shuai Yang, Yiming Luo +4 more
This paper presents RepWAM, a new representation-centric world action model that uses semantic visual-action tokenization to improve robot control.
Shiyuan Yang, Borong Zhang, Jizheng Zhang, Zhijia Tao +4 more
The paper introduces FabriVLA, a lightweight Vision-Language-Action model that achieves strong performance on the Meta-World MT50 benchmark using a compact 1B scale VLM backbone and a flow-matching ac…
Qihang Zhang, Lin Li, Luyao Zhang, Shuai Yang +25 more
This paper introduces LingBot-VA 2.0, a video-action foundation model designed for embodiment, with semantic visual-action tokenization, causal pretraining, sparse MoE backbone, and enhanced asynchron…
Yue Peng, Yongzhe Zhao, Artur Habuda, Khuyen Pham +4 more
Proposed G$^3$VLA, a camera-aware geometric module for pretrained vision-language-action models, injecting calibrated structure into visual tokens using intrinsic-conditioned ray embeddings, projectiv…
Ning Cheng, Jinan Xu, Wanlin Li, Yangzhi Chen +4 more
This paper introduces τ, a touch-augmented Vision-Language-Action framework that learns an action-conditioned spatiotemporal tactile representation from future visual supervision and fuses it with vis…
Lingyi Hong, Weiren Cai, Jiayu Chen, Xinyu Zhou +6 more
This paper presents SimpleVLN, an architecture-preserving framework for Vision-and-Language Navigation (VLN) that unlocks the potential of native multimodal large language models (MLLMs) without task-…
Jin Yao, Dhruva Dixith Kurra, Tom Lampo, Zezhou Cheng +2 more
The paper introduces VLGA, a vision-language-action model that grounds its actions in the 3D world by incorporating geometry as a fourth modality.
Zefu Lin, Rongxu Cui, Junjia Xu, Xiaojuan Jin +3 more
The paper presents World Pilot, a Vision-Language-Action framework that augments VLA models with priors from a World-Action Model to improve performance on manipulation tasks.
Jisang Han, Seonghu Jeon, Jaewoo Jung, René Zurbrügg +6 more
The paper proposes the Geometric Action Model (GAM), a language-conditioned manipulation policy that uses a pretrained geometric foundation model to reason about 3D geometry and perform contact-rich m…
Yuefeng Peng, Mingzhe Li, Kejing Xia, Renhao Zhang +1 more
This paper presents the first systematic study of membership inference attacks (MIAs) against Vision-Language-Action (VLA) models, demonstrating that these models are highly vulnerable to privacy brea…
Qiuyue Wang, Mingsheng Li, Jian Guan, Jinhui Ye +36 more
Qwen-VLA introduces a unified embodied foundation model that extends vision-language understanding to continuous action generation, enabling robust, multi-task generalization across diverse robotic ta…
Jiajun Cao, Kai Tang, Bohua Zhang, Zezhong Qian +16 more
This paper presents a closed-loop framework called LAW that integrates language reasoning, action generation, and world modeling for long-horizon manipulation tasks.