GST-VLA: Structured Gaussian Spatial Tokens for 3D Depth-Aware Vision-Language-Action Models
Yeungnam University（延世大学）；Korea Advanced Institute of Science and Technology, KAIST（韩国科学技术院）
Md Selim Sarowar, Omer Tariq, Sungho Kim
备注：
The results presented in this paper are preliminary. Please note that the experiments are currently ongoing, and the final data is subject to change upon the completion of the study. All ideas, results, methods, and any content herein are the sole property of the authors
VLA models encode visual observations as 2D patch tokens with no intrinsic geometric structure. We introduce GST-VLA with two contributions. First, the Gaussian Spatial Tokenizer (GST) converts frozen dense depth and frozen semantic patch features into $N_g{=}128$ anisotropic 3D Gaussian primitives, each parameterized by a metric residual mean $μ\in \mathbb{R}^3$, log-scale covariance $\log σ\in \mathbb{R}^3$, and learned opacity $α\in (0,1)$. The covariance eigenstructure encodes local surface orientation, and opacity provides per-primitive geometric confidence, both inaccessible from scalar depth. Spatial attention pooling with learned queries concentrates the fixed token budget on geometrically salient regions rather than distributing uniformly. Second, 3D Depth-Aware Chain-of-Thought (DA-CoT) reasoning supervises four structured intermediate spatial thoughts, covering 3D object grounding, grasp affordance contact geometry, pairwise metric distances, and coarse SE(3) waypoints, as explicit generation targets in the training loss. A cross-attention sublayer at every VLM transformer block provides direct access to the raw 256-primitive Gaussian field during DA-CoT generation. A 300M-parameter flow-matching action expert with mixture-of-experts feedforward sublayers decodes 7-DoF delta action chunks via conditional ODE integration, conditioned on both VLM hidden states and DA-CoT outputs through dual cross-attention. Trained with composite $\mathcal{L}_\mathrm{flow} + \mathcal{L}_\mathrm{CoT} + \mathcal{L}_\mathrm{depth}$ across three progressive stages, GST-VLA achieves 96.4% on LIBERO (+2.0%), and 80.2% on SimplerEnv (+5.4%). Ablations isolate the contribution of each GST component, each DA-CoT thought, and each training stage, confirming independent and synergistic gains concentrated on precision demanding tasks.
Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inference
School of AI, Shanghai Jiao Tong University（上海交通大学人工智能学院）；University of Science and Technology of China（中国科学技术大学）；Harbin Institute of Technology (Shenzhen)（哈尔滨工业大学（深圳））；BAAI（北京人工智能研究院）
SA-VLA: State-aware tokenizer for improving Vision-Language-Action Models' performance
Shanghai Jiao Tong University（上海交通大学）；East China University of Science and Technology（东华大学）；Hong Kong Polytechnic University（香港理工大学）；Xi’an University of Electronic Science and Technology（西安电子科技大学）
TokenPB: Enabling VLA Action Self-Verification via VLM Token Probability Bucketing
Key Laboratory of Data Engineering（数据工程与知识工程重点实验室）；Tsinghua University（清华大学）；Beijing University of Posts（北京邮电大学）；School of Information, Renmin University of China（中国人民大学信息学院）
SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration
Tsinghua University（清华大学）；The Chinese University of Hong Kong（香港中文大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Beihang University（北京航空航天大学）
ETA-VLA: Efficient Token Adaptation via Temporal Fusion and Intra-LLM Sparsification for Vision-Language-Action Models
Bosch Corporate Research, Bosch (China) Investment Ltd.（博世企业研究，博世（中国）投资有限公司）；School of Communication and Information Engineering, Shanghai University（上海大学通信与信息工程学院）
DFM-VLA: Iterative Action Refinement for Robot Manipulation via Discrete Flow Matching
The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；Harbin Institute of Technology（哈尔滨工业大学）；ShanghaiTech University（上海科技大学）；Shanghai Institute of Technical Physics, CAS（中国科学院上海技术物理研究所）
Think Proprioceptively: State-Grounded Visual Token Selection for VLA Policies
The Hong Kong Polytechnic University（香港理工大学）；Eastern Institute of Technology（东区技术研究所）；Great Bay University（大湾大学）；Northeast Forestry University（东北林业大学）；National University of Singapore（新加坡国立大学）
Token Expand-Merge: Training-Free Token Compression for Vision-Language-Action Models
College of Science and Engineering, Hamad Bin Khalifa University（1 科学与工程学院，哈马德·本·卡西姆大学）；Mohamed bin Zayed University of Artificial Intelligence（2 摩萨·本·扎耶德人工智能大学）；College of Computer Science and Technology, Zhejiang University（3 计算机科学与技术学院，浙江大学）
VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching
School of Computer Science, University of Sydney（悉尼大学计算机科学学院）；John Hopcropt Center for Computer Science, Shanghai Jiao Tong University（上海交通大学约翰·霍普克罗夫特计算机科学中心）
2D or 3D: Who Governs Salience in VLA Models? -- Tri-Stage Token Pruning Framework with Modality Salience Awareness
School of Computer Science, Peking University（北京大学计算机科学学院）；ZTE Corporation（中兴通讯）；School of Artificial Intelligence, Beijing Normal University（北京师范大学人工智能学院）；School of Computer Science, China University of Geosciences (Wuhan)（中国地质大学（武汉）计算机科学学院）；School of Electronics Engineering and Computer Science, Peking University（北京大学电子工程与计算机科学学院）
Compressor-VLA: Instruction-Guided Visual Token Compression for Efficient Robotic Manipulation
School of Information Science and Technology, Beijing University of Technology（信息科学与技术学院，北京理工大学）；LiAuto Inc.（LiAuto公司）；Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing University of Technology（计算智能与智能系统北京市重点实验室，北京理工大学）
One Token Per Frame: Reconsidering Visual Bandwidth in World Models for VLA Policy
Zhejiang University（浙江大学）；Central South University（中南大学）；Harbin Institute of Technology（哈尔滨工业大学）；Embodied Intelligence General Platform Laboratory, Chery Auto（奇瑞汽车 embodied intelligence 通用平台实验室）；E-surfing Digital Life Technology Co., Ltd., China Telecom（亿联数字生活技术有限公司，中国电信）
CoWorld-VLA: Thinking in a Multi-Expert World Model for Autonomous Driving
Afari Intelligent Drive（Afari智能驾驶公司）；University of Electronic Science and Technology of China（电子科技大学）；Shanghai Jiao Tong University（上海交通大学）；Beijing University Of Posts and Telecommunications（北京邮电大学）；Tianjin University（天津大学）
DTP: A Simple yet Effective Distracting Token Pruning Framework for Vision-Language Action Models
Australian National University（澳大利亚国立大学）；University of California, San Diego（加州大学圣地亚哥分校）；Chinese Academy of Sciences（中国科学院）；Beijing Institute of Graphic Communication（北京印刷学院）；University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校）；Baidu Search（百度搜索）；Bytedance（字节跳动）
eRLT: Efficient VLA Reinforcement Learning via Action-Relevant Token Routing
Southern University of Science and Technology（南方科技大学）；Beijing Zhongguancun Academy（北京中关村学院）；Samsung Robotics eXperience（三星机器人体验中心）；Wuhan University（武汉大学）；Sun Yat-sen University（中山大学）
E0: Enhancing Generalization and Fine-Grained Control in VLA Models via Tweedie Discrete Diffusion
Sun Yat-sen University（中山大学）；Guangdong Key Laboratory of Big Data Analysis（广东大数据分析与处理重点实验室）；X-Era AI Lab（X-Era人工智能实验室）；Guangdong University of Technology（广东工业大学）
FD-VLA: Force-Distilled Vision-Language-Action Model for Contact-Rich Manipulation
Advanced Robotics Centre, National University of Singapore（新加坡国立大学先进机器人中心）；School of Electrical & Electronic Engineering, Nanyang Technological University（南洋理工大学电子与电气工程学院）；College of Information Science and Technology, Eastern Institute of Technology（东部技术学院信息科学与技术学院）；John A. Paulson School of Engineering and Applied Sciences, Harvard University（哈佛大学约翰·A·保罗森工程与应用科学学院）；Advanced Robotics Centre at National University of Singapore（新加坡国立大学先进机器人中心）；Singapore Institute of Manufacturing Technology, Agency for Science, Technology and Research (A*STAR)（新加坡制造技术研究所，科技研究局（A*STAR））
VECTOR-Drive: Tightly Coupled Vision-Language and Trajectory Expert Routing for End-to-End Autonomous Driving
College of Automotive Engineering, Jilin University（吉林大学汽车工程学院）；The National Key Laboratory of Automotive Chassis Integration and Bionics, Jilin University（吉林大学汽车底盘集成与生物力学国家级重点实验室）；ReeFocus AI Technology（ReeFocus人工智能技术）
The Better You Learn, The Smarter You Prune: Towards Efficient Vision-language-action Models via Differentiable Token Pruning
LiAuto Inc.（LiAuto公司）；School of Vehicle and Mobility, Tsinghua University（清华大学车辆与移动学院）；Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）
KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models
School of Computer Science, Peking University, Beijing, China（北京大学计算机科学学院）；School of Computer Science, China University of Geosciences (Wuhan), Wuhan, China（中国地质大学（武汉）计算机科学学院）
M2Tok: Multi-head Multi-codebook Discrete Action Tokenization for Vision-Language-Action Models
The Hong Kong Polytechnic University（香港理工大学）；Shanghai AI Laboratory（上海人工智能实验室）；The University of Hong Kong（香港大学）；Shanghai Jiao Tong University（上海交通大学）；Hong Kong University of Science and Technology（香港科技大学）
BFA++: Hierarchical Best-Feature-Aware Token Prune for Multi-View Vision Language Action Model
Institute of Software, Chinese Academy of Sciences（中国科学院软件研究所）；University of Chinese Academy of Sciences（中国科学院大学）；LimX Dynamic（LimX动态）；Nanjing University（南京大学）；Nanyang Technological University（南洋理工大学）；Zhejiang University（浙江大学）
