Efficient VLA Model Inference

VLA: Vision-Language-Action

This topic studies how to run vision-language-action models fast enough for responsive, reliable robot control, including on edge devices. Common approaches prune or cache visual tokens, quantize or distill models, reduce action-generation steps, and overlap inference with observation and execution. Methods are tested on manipulation benchmarks such as LIBERO and RoboTwin, as well as real robots, using measures like latency, control frequency, memory, energy, and task success. The main tension is that faster inference can undermine closed-loop reliability or motion quality, so gains must be assessed in full task execution rather than by model speed alone.

Latest papers

190 papers56 papers in the last four weeks, up 115% on the four weeks before. 0.3% of all new papers.

GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI Agents
Oct 27, 2026Bingchao Wang, Yongcheng Jing, Shunyu Liu +5Mobile GUI Agents Mobile GUI Automation
Shanghai Jiao Tong University · Nanyang Technological University · Alibaba Cloud Computing
ASPEN: Adaptive Semantic Reasoning via Command Persistence Prediction for Efficient Navigation VLAs
Oct 26, 2026Vieri N. Akustia, Bin HuEfficient VLA Model Inference Vision-Language Navigation
Department of Electrical and Computer Engineering, University of Houston, Houston, TX 77004, USA · Departement of Engineering Technology, University of Houston, Houston, TX 77004, USA
VLA-HWBench: Where the Time Goes When VLA Models Run on Edge Hardware
Oct 26, 2026Borui Xu, Hongshi Tan, Wenqi Jiang +4Efficient VLA Model Inference On-Device Inference
National University of Singapore
TRIAGE: Efficient Vision-Language-Action Inference Needs Both Reuse and Pruning
Oct 25, 2026Sooyoung Kwon, Carlee Joe-Wong, Kyunghan LeeEfficient VLA Model Inference Visual Token Pruning
Dept. of ECE, Seoul National University. · Dept. of ECE, Carnegie Mellon University.
Power-of-Two Leaky Integrate-and-Fire Neurons for Event-Driven Vision-Language-Action Inference
Oct 24, 2026Rui Liu, He Lv, Ziyao Zhang +2Efficient VLA Model Inference LIF Neurons
Sivy Limited, Hangzhou, China
Coupling Execution Horizon and Sampling Depth in Flow-Matching VLA Policies
Oct 23, 2026Jiabing Yang, Yuan Xu, Yixiang Chen +13Flow-Based VLA Models Efficient VLA Model Inference
School of Artificial Intelligence, University of Chinese Academy of Sciences. · New Laboratory of Pattern Recognition (NLPR), Institute of Automation, Chinese Academy of Sciences. · FiveAges. +2
When One Step Is Enough: An Empirical Study of Euler Integration in Flow-Matching Policies
Oct 22, 2026Shivakanth Sujit, Luca Nunziante, Rousslan Fernand Julien Dossa +1Flow Matching Efficient VLA Model Inference
Araya Inc., Tokyo
AGQ: Action-Guided Ultra-Low-Bit Quantization for Vision-Language-Action Models
Oct 21, 2026Xuhong Qiu, Zhuoyu Wu, Qiawei Zheng +3VLM Quantization Efficient VLA Model Inference
School of Computer Science and Engineering, Southeast University, Nanjing, China · School of Information Technology, Monash University, Subang Jaya, Malaysia · Department of Electronic & Computer Engineering, Hong Kong University of Science & Technology, Hong Kong, China +1
PACT-VLA: Speculative Action-Chunk Verification with Pre-Commit Observations
Oct 20, 2026Yixiang Zhu, Ruilong Wu, Yonghao Chen +3Vision-Language-Action Models Robustness of VLA Models
The Hong Kong University of Science and Technology (Guangzhou) · One Robotics
SPAR: Structural Pruning and Action Recovery for Vision-Language-Action Models
Oct 19, 2026Shenli Zheng, Zhou Fang, Yuan Du +2Efficient VLA Model Inference LVLM Pruning
Nanjing University, Nanjing, China · Nanjing University of Information Science and Technology, Nanjing, China
PaceVLA: Adaptive Temporal Action Quantization in Vision-Language-Action Models
Oct 19, 2026Hojin Jeon, Jimin Lee, Taekwan Kim +6Efficient VLA Model Inference Vision-Language-Action Models
KAIST · Korea University · RLWRLD +1
VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse
Oct 18, 2026Sicheng He, Yi Pan, Qiuting Yu +10Efficient VLA Model Inference Flow-Based VLA Models
Zhejiang University
Scene-Adaptive VLA: Efficient Autonomous Driving via Dynamic Layer Routing
Oct 17, 2026Yujia Yang, Quan Yuan, Jiawei Fan +5Efficient VLA Model Inference VLMs for Autonomous Driving
State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications
FlashRT-PAI: Fast Kernels Are Not Fast Pipelines for Physical AI Inference
Oct 17, 2026Liang Su, David Chen, Jason Li +2Efficient Neural Network Inference Efficient VLA Model Inference
AnyDevice AI & Alps Lab
MambaAct: Efficient Mamba Action Experts for Flow-Matching Vision-Language-Action Models
Oct 16, 2026Khuyen Pham, Tran Nguyen Le, Edoardo Fazzari +1Efficient VLA Model Inference Vision-Language-Action Models
MBZUAI · Technical University of Denmark
FastVLA: Task-Conditioned and Temporally Adaptive Visual Token Compression for Efficient Vision-Language-Action Models
Oct 15, 2026Jiuyi Xu, Chengyuan Qian, Zhengzhong Tu +2Efficient VLA Model Inference Visual Token Compression
Colorado School of Mines · Texas A&M University · Microsoft AI
Fast-VLA: You Only Think Once for Robot Manipulation
Oct 15, 2026Kehong Gong, Guanli Hou, Xiaoxu Zheng +6Vision-Language-Action Models Closed-Loop Control of VLA Models
Huawei Technologies Co., Ltd · Huawei Central Research Institute
LTM: Latent Transition Based In-Context Memory for Vision-Language-Action Models
Oct 15, 2026Qixin Sun, WenLong Gao, Xiansheng Chen +3Memory-Augmented VLMs ICL for Robotics
Beihang University · BAAI
When to Ask for Help: A Decision-Theoretic Account and Measurement Standard for Slow-Module Scheduling in Dual-System Vision-Language-Action Models
Oct 15, 2026Shilin ZhangVision-Language-Action Models Efficient VLA Model Inference
North China University of Technology
Causal Analysis of Robot Foundation Models Enables Efficiency Gains
Oct 14, 2026Aiden Swann, Mathilde Papillon, Mozes Jacobs +13Vision-Language-Action Models Efficient VLA Model Inference
•Goodfire · aStanford University · bUniversity of California, Santa Barbara +1
Accelerating VLA Inference with Perturbation Informed Execution
Oct 14, 2026Jehun Lee, Hajun Lee, Jimin Ryu +4Efficient VLA Model Inference Vision-Language-Action Models
Seoul National University
TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency
Oct 13, 2026Jack Lu, Yurong You, Boyi Li +8Reasoning for VLA Models VLMs for Robotics
NVIDIA · New York University · University of California, Berkeley +2
$Δ$ Flow Matching for Robot Foundation Models
Oct 11, 2026Kuangji Zuo, Tuo An, Xinying Guo +9Flow Matching for Action Generation Robot Foundation Models
MARS Lab, Nanyang Technological University
Recovering Low-Bit Failures in Vision-Language-Action Policies via Quantization-Conditioned Reinforcement Learning
Oct 11, 2026Seongmin Park, Jaewon Lee, Yujun Heo +4VLM Quantization Efficient VLA Model Inference
Hanyang University, Republic of Korea · LG Electronics, Republic of Korea
Barrier-Constrained Guidance For Real-time Asynchronous VLA Execution
Oct 10, 2026Wenbo Yu, Jifan Li, Wenke Xia +5Inference-Time Guidance Efficient VLA Model Inference
Beijing Forestry University · Beijing Innovation Center of Humanoid Robotics · Gaoling School of Artificial Intelligence, Renmin University of China
Memory Without Curation: Learning Multi-Stream Latent Memory for Robot Policies
Oct 9, 2026Yilin Wu, Rutav Shah, Shuaiyi Huang +5Memory-Augmented VLMs Visuomotor Policy Learning
Amazon FAR · Carnegie Mellon University · The University of Texas at Austin +1
REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models
Oct 8, 2026Houlong Xiong, Zhenqi Qiu, Zechen Wang +8Robotic Control Vision-Language-Action Models
Shanghai Jiao Tong University · PrimeBot · ShanghaiTech University +2
FastJEV: Understanding Redundancy for Compact JEV Inference
Oct 8, 2026Jie Ma, Jie Gao, Yihang Liu +5Efficient Multimodal Inference Efficient VLA Model Inference
Xiamen University
SpikingVLA: Asynchronous Spiking Vision-Language-Action Models
Oct 7, 2026Jingya Wang, Dehao Zhang, Shuai Wang +3Efficient VLA Model Inference Spiking Neural Networks
University of Electronic Science and Technology of China · Shenzhen Loop Area Institute · The Chinese University of Hong Kong (Shenzhen)
CARE: Certifying Acceleration for Vision-Language-Action Inference
Oct 6, 2026Rui Liu, Tong Zheng, Jindong Gu +1Efficient VLA Model Inference Vision-Language-Action Models
University of Maryland, College Park · University of Oxford · Google

Oct 27, 2026J/K move · Enter open · S save

GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI Agents

Bingchao Wang, Yongcheng Jing, Shunyu Liu, Chengyu Wang+4

Shanghai Jiao Tong University · Nanyang Technological University · Alibaba Cloud Computing

Mobile GUI Agents Mobile GUI Automation Efficient VLA Model Inference Joint-Embedding Predictive Architecture Action Representation Learning

Open paper AskPDF arXiv

Mobile GUI agents are typically fine-tuned 7B-scale vision-language models (VLMs) that generate each action as text. However, these VLMs decode every action token by token, and even agents that score candidate actions keep a large language model or VLM in the loop, making them slow and memory-hungry. To address these issues, we propose GUI-JEPA, to our knowledge the first decoder-free GUI agent policy built on a Joint-Embedding Predictive Architecture (JEPA), which selects actions in the latent space of a JEPA encoder pretrained on unlabelled GUI screenshots. Its action- and region-conditioned latent predictor turns the screen latent into one latent per candidate action, letting lightweight heads score all candidates against the instruction in a single batched pass, without decoding a token. We further show that the scorer relies on this latent pathway and that the predictor is trained for the decision itself: a randomly initialised predictor does just as well as one pretrained on screen dynamics. On AndroidControl, GUI-JEPA is 2.8-4.0x faster on the same GPU than a Qwen2.5-VL-7B agent we fine-tune with the same inputs, while reaching 86-92% of its in-distribution accuracy with 6% of its parameters. Code is available at https://github.com/bcwang-sjtu/GUI-JEPA