RL for VLMs

RL: Reinforcement Learning · VLM: Vision-Language Model

This field studies how reinforcement learning can improve models that interpret images or video alongside language, and sometimes take actions in visual environments. Common approaches include group-relative policy optimization, PPO, and reinforcement learning from verifiable rewards, with feedback based on answer correctness, visual evidence, intermediate reasoning, preferences, or tool use. Applications range from image and video question answering, spatial reasoning, and medical analysis to navigation and robotic manipulation. A central challenge is ensuring that better scores reflect genuine use of visual evidence rather than language shortcuts, while designing reliable rewards and efficient training that generalize beyond the training setting.

Latest papers

476 papers68 papers in the last four weeks, up 94% on the four weeks before. 0.4% of all new papers.

Aligning Medical Vision Language Models at the Capability Boundary with Entropy Adaptive Rubric Grounding
Oct 29, 2026Hongyu Bai, Jiaqi Zeng, Chunzheng ZhuMedical VLMs Reasoning for VLMs
Shanghai Jiao Tong University Shanghai, China · Hunan University Changsha, China
From Reasoning to Action: Consequence-Grounded Alignment for Vision-Language-Action Models
Oct 29, 2026Yixue Zhang, Wenke Xia, Kun Wu +6VLMs for Robotics Reasoning for VLA Models
Peking University, Beijing, China. · Beijing Innovation Center of Humanoid Robotics, Beijing, China. · Renmin University of China, Beijing, China. +1
Timbre: Reliability-aware harmonic consensus for label-free multimodal test-time rexinforcement learning
Oct 29, 2026Haotian Lyu, Haoyu Wang, Ru Wang +4Test-Time Adaptation of VLMs Test-Time Adaptation
University of Melbourne · Mohamed bin Zayed University of Artificial Intelligence · The University of Tokyo +1
Grounding Multimodal Reasoning with Evidence-Ablated Negatives
Oct 29, 2026Chenxi Liu, Tianyi Xiong, Yanshuo Chen +1Multimodal Reasoning RL for VLMs
Department of Computer Science, University of Maryland, College Park, College Park, MD 20742
ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning
Oct 27, 2026Chaehyun Kim, Chi-Pin Huang, Jisoo Kim +4Reasoning for VLA Models Robot Learning from Video
NVIDIA · KAIST
PerceptRL: Making Visual Perception Rewardable in MLLMs from Answers to Interactions
Oct 27, 2026Tongkun Guan, Zhibo Yang, Mingkun Yang +3Vision-Language Models RL for VLMs
MoE Key Laboratory of Artificial Intelligence, AI Institute, School of Computer Science, Shanghai Jiao Tong University, Shanghai 200240, China · Qwen
VLA-Stem: A Lightweight Safety Mechanism for Vision-Language-Action Models via Reinforcement Learning
Oct 27, 2026Bojun Zou, Wenhao Yu, Jiaohao Li +6Vision-Language-Action Models Robot Collision Avoidance
University of Science and Technology of China (USTC) · Peking University (PKU) · University of Macau (UM)
Vision-Language Agents Can Learn in Their Own Latent Worlds
Oct 27, 2026Haoxi Li, Sikai Bai, Jianfei Ma +5RL for VLMs World Models for VLA Models
Department of Computer Science and Engineering, HKUST, Hong Kong, China · School of Computing, National University of Singapore, Singapore · Department of Computer Science, University of Southern California, Los Angeles, USA
GRASP: Group-Adaptive Hybrid RL-SFT Post-Training for Medical VQA without Reasoning Annotations
Oct 25, 2026Wanqi Dong, Wanyue Dong, Jingze Ge +1Medical VQA Medical VLMs
National University of Singapore · Shandong Provincial Daizhuang Hospital
Of Chains and Gains: What Reinforcement Learning changes on Reasoning-based Image Quality Assessment
Oct 23, 2026Yixuan Li, Yuxuan Jiang, Hanwei Zhu +4Image Quality Assessment Reinforcement Learning with Verifiable Rewards
University of Bristol, UK · Jiangxi University of Finance and Economics, China · China University of Mining and Technology, China +1
When to Stop Penalizing Actions? Martingale GRPO for VLA Reinforcement Learning
Oct 23, 2026Zongyao Zhao, Cheng-Jie Zhao, Ruihang Jiang +3Policy Optimization Group Relative Policy Optimization
The University of Hong Kong · Tsinghua University
From Visual Affinity to Learning Credit in Multimodal Reinforcement Learning
Oct 23, 2026Yuhang Han, Dingning Liu, Yujie Chen +7RL for VLMs Step-Level Credit Assignment in RL
The Hong Kong University of Science and Technology (Guangzhou) · Shanghai Jiao Tong University · Shadoweave +2
PEAR: Adaptive Advantage Reshaping for Perception and Exploration in Large Vision-Language Models
Oct 23, 2026Junxin Wang, Yingfeng Luo, Xiangchen Song +12Reasoning for VLMs Multimodal Reasoning
School of Computer Science and Engineering, Northeastern University, Shenyang, China · University of Michigan, Ann Arbor, US · Kunming University of Science and Technology, Kunming, China +1
Where Do Long Captions Fail? Position-Aware Diagnosis and Reinforcement Learning for Detailed Image Captioning
Oct 22, 2026Yuanze Hu, Zhichao Yang, Junwei Jing +9Image Captioning Faithfulness in VLMs
Beihang University · Kuaishou Technology · Fudan University +1
RECAP: Looking Once Is Not Enough for Vision-Language Reasoning
Oct 22, 2026Zhaolu Kang, Tailong Luo, Chenxin Li +13Reasoning for VLMs Visual Evidence Sensitivity in VLMs
Peking University · Chengdu Minto Tech
DRILL: Training World Models to Improve Policies, Not Predict Pixels
Oct 22, 2026Zhaolu Kang, Tailong Luo, Lei Wei +9World Models for VLA Models World Model Learning
Hunyuan, Tencent · Peking University
Edit, Don't Imitate: Minimal-Edit On-Policy Distillation for Multimodal Reasoning
Oct 21, 2026Wen Ye, Ailin Deng, Zirui Wang +8Reasoning for VLMs Vision-Language Models
Meta Superintelligence Labs · University of Southern California · University of California, Berkeley +2
AtomicRL: Atomic Reinforcement Learning for Long-Horizon Robotic Manipulation
Oct 21, 2026Likui Zhang, Zhe Zhu, Zhihao Zhan +7RL for Robotics Long-Horizon Robotic Manipulation
Sun Yat-sen University · JD Explore Academy · The Hong Kong Polytechnic University
Reinforcement Learning with Visually Attributable On-Policy Self-Distillation
Oct 19, 2026Shiao Xie, Siyu Chen, Jianwei Lv +2RL for VLMs On-Policy Self-Distillation
Baidu Inc., China
Replan for Error Recovery: Training Embodied VLMs through Verifiable Interaction with Action-Grouped Turn-GSPO
Oct 19, 2026Zhuoran Wang, Keqing Fan, Yang Li +6VLMs for Robotics Robot Task Planning
School of Information, Renmin University of China · Key Laboratory of Data Engineering and Knowledge Engineering, Beijing, China · Beihang University +2
Revealing and Bridging the Physical Executability Gap in Vision-Language Navigation
Oct 19, 2026Gang He, Zhile Cheng, Kepeng Xu +2Vision-Language Navigation RL for VLMs
Xidian University · Dartmouth College
Zephyrus-R1: Agentic Reinforcement Learning for Weather Reporting
Oct 19, 2026Fengfei Yu, Marshall Fisher, Isadora White +13RL for VLMs Reinforcement Learning with Verifiable Rewards
University of California, San Diego
AeroDASH: Asynchronous Vision-Language-Action Control for Continuous UAV Flight
Oct 18, 2026Zhe Liu, Yuxuan Zheng, Jinzhe Qin +2UAV Navigation RL for VLMs
Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China · Department of Mechanical and Energy Engineering, Southern University of Science and Technology, Shenzhen, China · Differential Robotics
SimpleVLN: Towards Simple and Frontier Vision-Language Navigation
Oct 16, 2026Lingyi Hong, Weiren Cai, Jiayu Chen +7Vision-Language Navigation Large Vision-Language Models
Fudan University · Nanyang Technological University · Lenovo Research
Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models
Oct 15, 2026Yuqi Xiong, Chunyi Peng, Zhipeng Xu +5VLM Distillation RL for VLMs
School of Computer Science and Engineering, Northeastern University, Shenyang, China · Alibaba Group, Hangzhou, China
Learning to Think and Explore in 3D World
Oct 15, 2026Yijiang Li, Bingyang Wang, Nuno VasconcelosVLM Reasoning 3D VQA
University of California San Diego · Georgia Institute of Technology
Enhancing Multimodal RLVR with Pairwise Token Contrast and Adaptive Trajectory Confidence
Oct 15, 2026Yiran Zhang, Hanbo Huang, Jielan Li +3RL for VLMs Multimodal Reasoning
Shanghai Jiao Tong University · Shanghai Artificial Intelligence Laboratory
X-Fly: Learning to Reach and Stop in Full-Chunk Aerial Vision-Language Navigation
Oct 15, 2026Ruibin Ju, Ruifeng Cai, Min Xu +1Aerial VLN Imitation Learning
ARIDGE, XPENG
Perception or Reasoning? Diagnosing and Tackling Reinforcement Learning Bottlenecks in VLMs
Oct 15, 2026Zeyuan Yang, Zhuohao Chen, Jiaben Chen +3Attention-Guided Distillation RL for VLMs
UMass Amherst · UIUC · MIT
RL Teaches VLMs Where to Attend but Not How to Perceive
Oct 14, 2026Ziqian Zhang, Junjie Liao, Zhining Zhang +1RL for VLMs Visual Evidence Sensitivity in VLMs
National University of Singapore · Beijing Normal University · Peking University +1

Oct 29, 2026J/K move · Enter open · S save

Aligning Medical Vision Language Models at the Capability Boundary with Entropy Adaptive Rubric Grounding

Hongyu Bai, Jiaqi Zeng, Chunzheng Zhu

Shanghai Jiao Tong University Shanghai, China · Hunan University Changsha, China

Medical VLMs Reasoning for VLMs RL for VLMs Preference Optimization Adaptive Reasoning in Language Models Medical Image Grounding

Open paper AskPDF arXiv

Medical diagnostic reasoning requires verifiable links between visual evidence and clinical knowledge, yet current medical vision language models are aligned with objectives that reward the wrong behavior. Uniform supervised fine tuning lets easy cases dominate updates while clinically critical boundary cases saturate prematurely. Sparse rewards based only on final-answer correctness cannot assess anatomical grounding or clinical logic, inviting reward hacking through answer shortcuts. We propose MedCBR, a two-stage alignment framework coupling boundary-aware preference optimization with rubric-grounded reinforcement learning. First, we construct counterfactual boundary preferences at the model's capability frontier using empirical pass rates and preference pairs that share reasoning prefixes but diverge at decisive diagnostic branches. Difficulty Weighted Boundary Preference Optimization (DBPO) concentrates learning on this calibrated boundary region. Second, we distill reusable clinical rubrics from boundary trajectories and activate them according to prediction entropy. Entropy Gated Grounded Policy Optimization (EGPO) rewards answer correctness, anatomical grounding, and reasoning fidelity while suppressing rubric noise on reliable samples. A reference policy anchored to the boundary-aligned model couples population-level boundary priors with instance-level uncertainty, making diagnostic reasoning more accurate and clinically grounded. Experiments on six medical benchmarks show consistent improvements over strong general, specialized, and reinforcement learning baselines, reaching 61.2% average accuracy with the largest gains on challenging differential diagnosis tasks. Ablations show super-additive gains: combining the two stages exceeds the sum of their individual improvements.