Reasoning for VLA Models

VLA: Vision-Language-Action

This area studies how vision-language-action agents can interpret instructions, track spatial and physical information, plan over time, and turn decisions into robot, navigation, or driving actions. Approaches include explicit text or visual chain-of-thought, 3D spatial representations, latent plans and world models, hierarchical planners with action controllers, and training with demonstrations, teacher-generated rationales, or reinforcement learning. Evaluation spans simulated and real-robot manipulation, vision-language navigation, and autonomous driving, with growing attention to long-horizon tasks, perturbations, and instruction changes. A central open question is whether reasoning actually causes better, more faithful actions—or merely accompanies behavior driven by learned priors—while added deliberation must be balanced against latency, robustness, and reliable evaluation of physical understanding.

Latest papers

113 papers24 papers in the last four weeks, up 118% on the four weeks before. 0.1% of all new papers.

Look Before You Leap: Distilling Tree Search into Action Evaluation for Frozen VLA Models
Jul 4, 2026Xinyi Xie, Zican Hu, Zhanyu Liu +7Value Function Estimation Test-Time Scaling
Nanjing University · Australian National University · Institute of Automation, Chinese Academy of Sciences +1
Reasoning-aware Speculative Decoding for Efficient Vision-Language-Action Models in Autonomous Driving
Jun 30, 2026Anh Dung Dinh, Simon Khan, Flora SalimEfficient VLM Inference Speculative Decoding
School of Computer Science and Engineering UNSW Sydney · Air Force Research Laboratory
Training Vision-Language-Action Models with Dense Embodied Chain-of-Thought Supervision
Jun 29, 2026Haoyang Li, Guanlin Li, Youhe Feng +9Cross-Embodiment Robot Learning Robotic Manipulation
1Renmin University of China · 2Zhipu AI
Position: Vision-Language-Action Models Cannot Be Verified to Perform Physical Reasoning
Jun 28, 2026Taozhao Chen, Ian Manchester, Huaming ChenVLM Evaluation Physical Reasoning
School of Electrical and Computer Engineering The University of Sydney Sydney, NSW, Australia · School of Aerospace, Mechanical and Mechatronic Engineering The University of Sydney Sydney, NSW, Australia
TAP-VLA: Tactile Annotation Prompting for Vision Language Action Models
Jun 27, 2026Mark Van der Merwe, Mohamad Louai Shehab, Jayjun Lee +4Vision-Language-Action Models Tactile Sensing for Robotic Manipulation
Robotics Department, University of Michigan · Computer Science and Engineering Department, University of Michigan
FOCA: Future-Oriented Conditioning for Data-Efficient Vision-Language-Action Adaptation
Jun 18, 2026Duc Minh Nguyen, Nghiem Tuong Diep, Binh Gia Nguyen +20VLM Adaptation Efficient VLA Models
VinRobotics, Vietnam · Center for AI Research, VinUniversity, Vietnam · University of Utah, USA +4
Tri-Info: Generalizable, Interpretable Failure Prediction for VLA Models via Information Theory
Jun 18, 2026Jinghan Yang, Yunchao Zhang, Wang Yuan +4Robot Failure Detection Vision-Language-Action Models
HKU Musketeers Foundation Institute of Data Science · InfoBodied AI Lab, The University of Hong Kong
ThinkingVLA: Interleaved Vision and Language Reasoning for Robotic Manipulation
Jun 16, 2026Tianyi Lu, Hui Zhang, Zijie Diao +8Long-Horizon Robotic Manipulation Language-Conditioned Robot Manipulation
Fudan University · Shanghai Innovation Institue · China Unicom +1
PearlVLA: Progressive Embodied Action-Plan Refinement in Latent Space
Jun 16, 2026Bochen Yang, Lianlei ShanEfficient VLA Models Latent Visual Reasoning
Imperial College London · Tsinghua University
Think Less, Act Early: Reinforced Latent Reasoning with Early Exit in Vision-Language-Action Models
Jun 13, 2026Dianqiao Lei, Lianlei ShanEfficient Inference Efficient VLA Model Inference
1Tsinghua University, Beijing, China.
RT-VLA: Real-Time Vision-Language-Action Models via Knowledge Distillation
Jun 12, 2026Xiangyu Huang, Zhenlin Hua, Han Zhou +2End-to-End Autonomous Driving Efficient VLM Inference
Carnegie Mellon University
VLADriveBench: Evaluating CoT-Action Relationship in VLA for Autonomous Driving
Jun 10, 2026Thach Nguyen, Danhua Guo, Tom Lampo +2Autonomous Driving Benchmarks CoT Faithfulness
Uber AV Labs
MPCoT: Reward-Guided Multi-Path Latent Reasoning for Test-Time Scalable Vision-Language-Action
Jun 4, 2026Boyang Zhang, Lianlei ShanVisuomotor Policy Learning Robotic Manipulation
Department of Electrical and Computer Engineering, Boston University, Boston, MA 02215, USA. · Department of Computer Science, Tsinghua University, Beijing 100084, China.
3DThinkVLA: Endowing Vision-Language-Action Models with Latent 3D Priors via 3D-Thinking-Guided Co-training
Jun 3, 2026Jiaxin Shi, Xidong Zhang, Fucai Zhu +33D Spatial Reasoning Robotic Manipulation
Shanghai Jiao Tong University · Daimon Robotics · Harbin Institute of Technology +4
Continuous Reasoning for Vision-Language-Action
May 29, 2026Yueh-Hua Wu, Tatsuya Matsushima, Kei OtaVision-Language-Action Models VLMs for Robotics
AIRoA
SpatialAct: Probing Spatial Reasoning-to-Action Capabilities of VLM Agents in 3D Scenes
May 29, 2026Tianhui Liu, Jie Feng, Zhiheng Zheng +6Vision-Language Agents Visual Spatial Reasoning
1The Hong Kong University of Science and Technology (Guangzhou) · 2Zhongguancun Academy · 3Tsinghua University +1
VisualThink-VLA: Visual Intermediate Reasoning for Effective and Low-Latency Vision-Language-Action Policies
May 28, 2026Mingjian Gao, Wenqiao Zhang, Yuqian Yuan +9Efficient VLM Inference Efficient VLA Model Inference
1Zhejiang University · 2Cornell University · 3National University of Singapore +1
ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models
May 28, 2026Ye Li, Huanan Liu, Kangye Ji +7Efficient VLM Inference Efficient VLA Model Inference
Tsinghua University · University of Illinois at Urbana-Champaign
AwareVLN: Reasoning with Self-awareness for Vision-Language Navigation
May 21, 2026Wenxuan Guo, Xiuwei Xu, Yichen Liu +7Vision-Language Navigation VLM Reasoning
Tsinghua University, China
Pelican-Unify 1.0: A Unified Embodied Intelligence Model for Understanding, Reasoning, Imagination and Action
May 14, 2026Yi Zhang, Yinda Chen, Che Liu +26Action-Conditioned Video Generation Unified Multimodal Models
Beijing Innovation Center of Humanoid Robotics (X-Humanoid) WFM System Group
Guide, Think, Act: Interactive Embodied Reasoning in Vision-Language-Action Models
May 13, 2026Yiran Ling, Qing Lian, Jinghang Li +6Robotic Control Robot Policy Learning
Faculty of Computing, Harbin Institute of Technology · International Digital Economy Academy (IDEA) · National Key Laboratory of Smart Farm Technologies and Systems +4
PACE: Policy-Native Adaptive Decision Timing for Long-Horizon Reasoning
May 11, 2026Chen Li, Zhantao Yang, Fangyi Chen +3Large Vision-Language Models Long-Horizon Planning
Carnegie Mellon University
MolmoAct2: Action Reasoning Models for Real-world Deployment
May 4, 2026Haoquan Fang, Jiafei Duan, Donovan Clay +26Embodied Reasoning Efficient VLA Models
1Allen Institute for AI · University of Washington · 3National University of Singapore +6
Being-H0.7: A Latent World-Action Model from Egocentric Videos
Apr 30, 2026Hao Luo, Wanpeng Zhang, Yicheng Feng +6Visuomotor Policy Learning Efficient VLA Models
BeingBeyond Team
Xiaomi OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation
Apr 20, 2026Jinghui Lu, Jiayi Guan, Zhijian Huang +47Trajectory Generation VLMs for Autonomous Driving
Unmasking the Illusion of Embodied Reasoning in Vision-Language-Action Models
Apr 20, 2026Haiweng Xu, Sipeng Zheng, Hao Luo +3VLM Evaluation VLM Reasoning
1Peking University · 3BeingBeyond · 2Tsinghua University
ReFineVLA: Multimodal Reasoning-Aware Generalist Robotic Policies via Teacher-Guided Fine-Tuning
Apr 20, 2026Tuan Van Vo, Tan Q. Nguyen, Khang Nguyen +5Robotic Manipulation VLM Reasoning
VinRobotics, Hanoi. Vietnam · Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), UAE. · University of Texas at Arlington, Texas, USA +4
World-Value-Action Model: Implicit Planning for Vision-Language-Action Systems
Apr 16, 2026Runze Li, Hongyin Zhang, Junxi Jin +5Vision-Language-Action Models Model-Based Planning
Westlake University Hangzhou, China · Nanjing University Suzhou Campus, 1520 Taihu Avenue, Suzhou, China
ProGAL-VLA: Grounded Alignment through Prospective Reasoning in Vision-Language-Action Models
Apr 10, 2026Nastaran Darabi, Amit Ranjan TrivediSpatial Grounding for VLA Models Vision-Language-Action Models
University of Illinois Chicago, IL, USA
Orion-Lite: Distilling LLM Reasoning into Efficient Vision-Only Driving Models
Apr 9, 2026Jing Gu, Niccolò Cavagnero, Gijs DubbelmanEnd-to-End Autonomous Driving VLM Distillation
Eindhoven University of Technology

Jul 4, 2026J/K move · Enter open · S save

Look Before You Leap: Distilling Tree Search into Action Evaluation for Frozen VLA Models

Xinyi Xie, Zican Hu, Zhanyu Liu, Yicheng Dong+6

Nanjing University · Australian National University · Institute of Automation, Chinese Academy of Sciences+1

Value Function Estimation Test-Time Scaling Vision-Language-Action Models VLMs for Robotics Reasoning for VLA Models

Open paper AskPDF arXiv

Vision-Language-Action (VLA) models acquire broad embodied capabilities through large-scale pretraining, yet their generalization remains far more fragile than that of LLMs and VLMs. The prevailing remedy, post-training via supervised fine-tuning or reinforcement learning, improves task-specific performance but narrows the generalist capability that makes pretraining valuable. We identify a key bottleneck: VLA failures stem not only from action generation but also from action evaluation. A diagnostic pass@k study confirms that frozen VLAs already contain competent behaviors in their output distribution, with overall success rates rising from 33% at pass@1 to 92% at pass@32. Inspired by this, we propose SVA (Search, Value, and Act), a simple framework that equips frozen VLA policies with long-term consequence awareness. SVA first uses Monte-Carlo tree search in simulation to sample and explore the VLA's output distribution under a finite search budget, collecting diverse trajectories annotated with empirical returns; this knowledge is then distilled into a lightweight Q-value model that predicts the expected consequence of candidate actions; at deployment, the frozen VLA proposes multiple candidates and the evaluator selects the one with the highest uncertainty-regularized Q-value, requiring no simulator access. By decoupling action proposal from consequence evaluation, SVA preserves the generalization capacity of the VLA backbone while improving task success. Across embodied benchmarks, SVA consistently improves generalization on unseen tasks and exhibits strong test-time scaling; across real-robot manipulation tasks, it improves average success from 42.8% to 55.6% (+12.8 points). Strikingly, SVA enables a 9B VLA to outperform a 27B VLA by 7 points at 27% lower inference latency, suggesting that scaling test-time evaluation is more cost-effective than scaling model size.