论文研究
arxiv.org ·
FIVE-VLA：基于循环动作记忆的快速高效自动驾驶 VLA 模型
该论文提出 FIVE-VLA，针对自动驾驶 VLA 模型参数量大、高分辨率图像处理低效、缺乏时序记忆的问题给出两项改进：一是高效视觉编码器，处理 448×896 图像仅生成 98 个 token（较现有方法少 5 倍以上），并完全跳过文本生成，直接单次预测轨迹；二是 Recurrent Action Memory（RAM）轻量模块，以历史动作 token 为条件提供时序上下文，用于超车、紧急制动等场景。模型仅 641M 参数，在 Bench2Drive 闭环基准上无交通违规完成路线比此前 SOTA VLA 多约 10%；在 NVIDIA Physical AI AV 数据集开环仿真中，单视角与四视角碰撞违规率分别比 SimLingo 低 10.2% 和 7.7%；A100 上约 30 fps、T4 上约 4 fps，较此前方法提速 8-30 倍。
AI推荐理由 arXiv 新提交论文（2026-09-16），作者与摘要信息完整，属可验证的一手研究。核心新信息：FIVE-VLA 用高效视觉编码器把 448×896 高分辨率图像压到仅 98 个 token（比现有方法少 5 倍以上），并跳过文本生成直接单次预测轨迹；提出 Recurrent Action Memory（RAM）轻量模块，用历史动作 token 提供时序上下文。仅 641M 参数，在 Bench2Drive 闭环基准上无违规完成路线比此前 SOTA VLA 多约 10%，在 NVIDIA Physical AI AV 数据集开环仿真中碰撞违规率比 SimLingo 低 10.2%（单视角）和 7.7%（四视角），A100 上约 30 fps、T4 上约 4 fps，较此前方法提速 8-30 倍。局限：结果均为论文自报的仿真/基准指标，尚无真实道路部署证据，需复现验证。
78
正文
Computer Science > Computer Vision and Pattern Recognition
arXiv:2609.18623 (cs)
[Submitted on 16 Sep 2026]
Title:FIVE-VLA: Fast and EffectIVE Autonomous Driving with Recurrent Action Memory
Authors:Kemal Oksuz, Alexandru Buburuzan, Yuhan Yao, Puneet K. Dokania
View a PDF of the paper titled FIVE-VLA: Fast and EffectIVE Autonomous Driving with Recurrent Action Memory, by Kemal Oksuz and Alexandru Buburuzan and Yuhan Yao and Puneet K. Dokania
View PDF
HTML (experimental)
Abstract:State-of-the-art vision-language-action models (VLA) for autonomous driving face critical limitations: excessive parameter counts, inefficient high-resolution image processing, and lack of temporal memory. We introduce Fast and EffectIVE VLA (FIVE-VLA) to address these through two key contributions. First, we employ an efficient vision encoder that processes high-resolution ($448 \times 896$) images while generating only 98 tokens, over $5\times$ fewer than existing approaches, and bypass text generation entirely for single-pass trajectory prediction. Second, we propose Recurrent Action Memory (RAM), a lightweight module that conditions action prediction on previous action tokens, providing temporal context critical for manoeuvres such as overtaking and emergency braking. With only 641M parameters, FIVE-VLA completes $\sim$10% more routes without traffic rule infractions than the previous state-of-the-art VLA on the challenging Bench2Drive closed-loop driving benchmark. Non-reactive open-loop simulation on the large-scale real-world NVIDIA Physical AI AV dataset shows 10.2% and 7.7% lower collision-violation rates than SimLingo in single- and four-view settings, respectively. Additionally, FIVE-VLA runs at $\sim$30 fps on an A100 and $\sim$4 fps on a T4 GPU (proxy to an edge device), representing an 8-30$\times$ speedup over previous methods.
Subjects:
Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
Cite as:
arXiv:2609.18623 [cs.CV]
(or
arXiv:2609.18623v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.18623
Focus to learn more
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Kemal Oksuz [view email]
[v1]
Wed, 16 Sep 2026 13:13:18 UTC (4,422 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled FIVE-VLA: Fast and EffectIVE Autonomous Driving with Recurrent Action Memory, by Kemal Oksuz and Alexandru Buburuzan and Yuhan Yao and Puneet K. Dokania
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CV
new
|
recent
| 2026-09
Change to browse by:
cs
cs.RO
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)