Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhang, Yisheng, Jia, Guoli, Hu, Haote, Zhao, Shanxu, Zhao, Kaikai, Sun, Long, Long, Xinwei, Tian, Kai, Jiang, Che, Liu, Zhaoxiang, Wang, Kai, Lian, Shiguo, Zhang, Kaiyan, Zhou, Bowen
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.27742
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917366433054720
author Zhang, Yisheng
Jia, Guoli
Hu, Haote
Zhao, Shanxu
Zhao, Kaikai
Sun, Long
Long, Xinwei
Tian, Kai
Jiang, Che
Liu, Zhaoxiang
Wang, Kai
Lian, Shiguo
Zhang, Kaiyan
Zhou, Bowen
author_facet Zhang, Yisheng
Jia, Guoli
Hu, Haote
Zhao, Shanxu
Zhao, Kaikai
Sun, Long
Long, Xinwei
Tian, Kai
Jiang, Che
Liu, Zhaoxiang
Wang, Kai
Lian, Shiguo
Zhang, Kaiyan
Zhou, Bowen
contents Vision-language agents that orchestrate specialized tools for image restoration (IR) have emerged as a promising method, yet most existing frameworks operate in a training-free manner. They rely on heuristic task scheduling and exhaustive tool traversal, resulting in sub-optimal restoration paths and prohibitive computational cost. We argue that the core bottleneck lies in the absence of a learned policy to make decision, as a vision-language model cannot efficiently handle degradation-aware task ordering and tool composition. To this end, we propose TIR-Agent, a trainable image restoration agent that performs a direct tool-calling policy through a two-stage training pipeline of supervised fine-tuning (SFT) followed by reinforcement learning (RL). Two key designs underpin effective RL training: (i) a random perturbation strategy applied to the SFT data, which broadens the policy's exploration over task schedules and tool compositions, and (ii) a multi-dimensional adaptive reward mechanism that dynamically re-weights heterogeneous image quality metrics to mitigate reward hacking. To support high-throughput, asynchronous GPU-based tool invocation during training, we further develop a globally shared model-call pool. Experiments on both in-domain and out-of-domain degradations show that TIR-Agent outperforms 12 baselines, including 6 all-in-one models, 3 training-free agents, and 3 proprietary models, and achieves over 2.5$\times$ inference speedup by eliminating redundant tool executions.
format Preprint
id arxiv_https___arxiv_org_abs_2603_27742
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TIR-Agent: Training an Explorative and Efficient Agent for Image Restoration
Zhang, Yisheng
Jia, Guoli
Hu, Haote
Zhao, Shanxu
Zhao, Kaikai
Sun, Long
Long, Xinwei
Tian, Kai
Jiang, Che
Liu, Zhaoxiang
Wang, Kai
Lian, Shiguo
Zhang, Kaiyan
Zhou, Bowen
Computer Vision and Pattern Recognition
Vision-language agents that orchestrate specialized tools for image restoration (IR) have emerged as a promising method, yet most existing frameworks operate in a training-free manner. They rely on heuristic task scheduling and exhaustive tool traversal, resulting in sub-optimal restoration paths and prohibitive computational cost. We argue that the core bottleneck lies in the absence of a learned policy to make decision, as a vision-language model cannot efficiently handle degradation-aware task ordering and tool composition. To this end, we propose TIR-Agent, a trainable image restoration agent that performs a direct tool-calling policy through a two-stage training pipeline of supervised fine-tuning (SFT) followed by reinforcement learning (RL). Two key designs underpin effective RL training: (i) a random perturbation strategy applied to the SFT data, which broadens the policy's exploration over task schedules and tool compositions, and (ii) a multi-dimensional adaptive reward mechanism that dynamically re-weights heterogeneous image quality metrics to mitigate reward hacking. To support high-throughput, asynchronous GPU-based tool invocation during training, we further develop a globally shared model-call pool. Experiments on both in-domain and out-of-domain degradations show that TIR-Agent outperforms 12 baselines, including 6 all-in-one models, 3 training-free agents, and 3 proprietary models, and achieves over 2.5$\times$ inference speedup by eliminating redundant tool executions.
title TIR-Agent: Training an Explorative and Efficient Agent for Image Restoration
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.27742