Saved in:
Bibliographic Details
Main Authors: Song, Zirui, Ouyang, Guangxian, Li, Mingzhe, Ji, Yuheng, Wang, Chenxi, Xu, Zixiang, Zhang, Zeyu, Zhang, Xiaoqing, Jiang, Qian, Chen, Zhenhao, Li, Zhongzhi, Yan, Rui, Chen, Xiuying
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.16517
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916756261437440
author Song, Zirui
Ouyang, Guangxian
Li, Mingzhe
Ji, Yuheng
Wang, Chenxi
Xu, Zixiang
Zhang, Zeyu
Zhang, Xiaoqing
Jiang, Qian
Chen, Zhenhao
Li, Zhongzhi
Yan, Rui
Chen, Xiuying
author_facet Song, Zirui
Ouyang, Guangxian
Li, Mingzhe
Ji, Yuheng
Wang, Chenxi
Xu, Zixiang
Zhang, Zeyu
Zhang, Xiaoqing
Jiang, Qian
Chen, Zhenhao
Li, Zhongzhi
Yan, Rui
Chen, Xiuying
contents Large Vision-Language Models (LVLMs) have recently advanced robotic manipulation by leveraging vision for scene perception and language for instruction following. However, existing methods rely heavily on costly human-annotated training datasets, which limits their generalization and causes them to struggle in out-of-domain (OOD) scenarios, reducing real-world adaptability. To address these challenges, we propose ManipLVM-R1, a novel reinforcement learning framework that replaces traditional supervision with Reinforcement Learning using Verifiable Rewards (RLVR). By directly optimizing for task-aligned outcomes, our method enhances generalization and physical reasoning while removing the dependence on costly annotations. Specifically, we design two rule-based reward functions targeting key robotic manipulation subtasks: an Affordance Perception Reward to enhance localization of interaction regions, and a Trajectory Match Reward to ensure the physical plausibility of action paths. These rewards provide immediate feedback and impose spatial-logical constraints, encouraging the model to go beyond shallow pattern matching and instead learn deeper, more systematic reasoning about physical interactions.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16517
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ManipLVM-R1: Reinforcement Learning for Reasoning in Embodied Manipulation with Large Vision-Language Models
Song, Zirui
Ouyang, Guangxian
Li, Mingzhe
Ji, Yuheng
Wang, Chenxi
Xu, Zixiang
Zhang, Zeyu
Zhang, Xiaoqing
Jiang, Qian
Chen, Zhenhao
Li, Zhongzhi
Yan, Rui
Chen, Xiuying
Robotics
Computer Vision and Pattern Recognition
Large Vision-Language Models (LVLMs) have recently advanced robotic manipulation by leveraging vision for scene perception and language for instruction following. However, existing methods rely heavily on costly human-annotated training datasets, which limits their generalization and causes them to struggle in out-of-domain (OOD) scenarios, reducing real-world adaptability. To address these challenges, we propose ManipLVM-R1, a novel reinforcement learning framework that replaces traditional supervision with Reinforcement Learning using Verifiable Rewards (RLVR). By directly optimizing for task-aligned outcomes, our method enhances generalization and physical reasoning while removing the dependence on costly annotations. Specifically, we design two rule-based reward functions targeting key robotic manipulation subtasks: an Affordance Perception Reward to enhance localization of interaction regions, and a Trajectory Match Reward to ensure the physical plausibility of action paths. These rewards provide immediate feedback and impose spatial-logical constraints, encouraging the model to go beyond shallow pattern matching and instead learn deeper, more systematic reasoning about physical interactions.
title ManipLVM-R1: Reinforcement Learning for Reasoning in Embodied Manipulation with Large Vision-Language Models
topic Robotics
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.16517