LENS: Learning to Segment Anything with Unified Reinforced Reasoning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhu, Lianghui, Ouyang, Bin, Zhang, Yuxuan, Cheng, Tianheng, Hu, Rui, Shen, Haocheng, Ran, Longjin, Chen, Xiaoxin, Yu, Li, Liu, Wenyu, Wang, Xinggang
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911272824471552
author Zhu, Lianghui
Ouyang, Bin
Zhang, Yuxuan
Cheng, Tianheng
Hu, Rui
Shen, Haocheng
Ran, Longjin
Chen, Xiaoxin
Yu, Li
Liu, Wenyu
Wang, Xinggang
author_facet Zhu, Lianghui
Ouyang, Bin
Zhang, Yuxuan
Cheng, Tianheng
Hu, Rui
Shen, Haocheng
Ran, Longjin
Chen, Xiaoxin
Yu, Li
Liu, Wenyu
Wang, Xinggang
contents Text-prompted image segmentation enables fine-grained visual understanding and is critical for applications such as human-computer interaction and robotics. However, existing supervised fine-tuning methods typically ignore explicit chain-of-thought (CoT) reasoning at test time, which limits their ability to generalize to unseen prompts and domains. To address this issue, we introduce LENS, a scalable reinforcement-learning framework that jointly optimizes the reasoning process and segmentation in an end-to-end manner. We propose unified reinforcement-learning rewards that span sentence-, box-, and segment-level cues, encouraging the model to generate informative CoT rationales while refining mask quality. Using a publicly available 3-billion-parameter vision-language model, i.e., Qwen2.5-VL-3B-Instruct, LENS achieves an average cIoU of 81.2% on the RefCOCO, RefCOCO+, and RefCOCOg benchmarks, outperforming the strong fine-tuned method, i.e., GLaMM, by up to 5.6%. These results demonstrate that RL-driven CoT reasoning significantly enhances text-prompted segmentation and offers a practical path toward more generalizable Segment Anything models (SAM). Code is available at https://github.com/hustvl/LENS.
format Preprint
id arxiv_https___arxiv_org_abs_2508_14153
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LENS: Learning to Segment Anything with Unified Reinforced Reasoning
Zhu, Lianghui
Ouyang, Bin
Zhang, Yuxuan
Cheng, Tianheng
Hu, Rui
Shen, Haocheng
Ran, Longjin
Chen, Xiaoxin
Yu, Li
Liu, Wenyu
Wang, Xinggang
Computer Vision and Pattern Recognition
Artificial Intelligence
Text-prompted image segmentation enables fine-grained visual understanding and is critical for applications such as human-computer interaction and robotics. However, existing supervised fine-tuning methods typically ignore explicit chain-of-thought (CoT) reasoning at test time, which limits their ability to generalize to unseen prompts and domains. To address this issue, we introduce LENS, a scalable reinforcement-learning framework that jointly optimizes the reasoning process and segmentation in an end-to-end manner. We propose unified reinforcement-learning rewards that span sentence-, box-, and segment-level cues, encouraging the model to generate informative CoT rationales while refining mask quality. Using a publicly available 3-billion-parameter vision-language model, i.e., Qwen2.5-VL-3B-Instruct, LENS achieves an average cIoU of 81.2% on the RefCOCO, RefCOCO+, and RefCOCOg benchmarks, outperforming the strong fine-tuned method, i.e., GLaMM, by up to 5.6%. These results demonstrate that RL-driven CoT reasoning significantly enhances text-prompted segmentation and offers a practical path toward more generalizable Segment Anything models (SAM). Code is available at https://github.com/hustvl/LENS.
title LENS: Learning to Segment Anything with Unified Reinforced Reasoning
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2508.14153