pith. sign in

arxiv: 2605.22208 · v1 · pith:UJBYAYSRnew · submitted 2026-05-21 · 💻 cs.CV

EvoIR-Agent: Self-Evolving Image Restoration Agentic System via Experience-Driven Learning

Kailin Zhuang , Jiawei Wu , Zhi Jin This is my paper

Pith reviewed 2026-05-22 07:45 UTC · model grok-4.3

classification 💻 cs.CV
keywords image restorationagentic systemmultimodal large language modelself-evolving mechanismexperience-driven learningtool selectiondegradation removal
0
0 comments X

The pith

A self-evolving hierarchical experience pool lets image restoration agents improve performance and efficiency while staying compatible with new tools.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Current MLLM-driven image restoration agents must choose between training-based methods that embed experience for speed but cannot adapt to new tools, and training-free methods that store experience explicitly for flexibility but still require heavy trial-and-error. The paper proposes EvoIR-Agent to first formulate the experience components of a training-free agent, then build a hierarchical experience pool that supplies coarse-to-fine guidance on tool choice and removal order. It adds a self-evolving mechanism that starts from scratch and updates the pool with accumulated records. This is intended to produce stronger results on coupled degradations. A sympathetic reader would care because the method aims to combine the efficiency of trained systems with the adaptability of explicit storage.

Core claim

EvoIR-Agent first systematically formulates the experience components of a training-free image restoration agent. Subsequently, a hierarchical experience pool is constructed, which enables coarse-to-fine guidance for diverse tools and removal orders. Furthermore, a self-evolving mechanism is introduced to update the pool from scratch using accumulated records, thereby greatly improving performance and efficiency.

What carries the argument

Hierarchical experience pool updated by a self-evolving mechanism that accumulates records to refine tool selection and removal-order decisions.

If this is right

  • Achieves a significant lead in full-reference metrics over state-of-the-art methods.
  • Yields a Pareto-optimal balance between performance and efficiency.
  • Preserves compatibility with new tools and degradations without retraining.
  • Reduces trial-and-error overhead that zero-shot planning otherwise incurs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same experience-pool structure could be tested in other MLLM agent domains that require sequential tool use.
  • Continuous record accumulation might allow the system to handle gradually shifting real-world degradation distributions.
  • The approach suggests a general pattern for making training-free agents improve over time without losing flexibility.

Load-bearing premise

The self-evolving mechanism can update the hierarchical experience pool from scratch using accumulated records to greatly improve performance and efficiency while preserving compatibility with new tools and degradations.

What would settle it

Run the agent on standard image-restoration benchmarks with held-out degradation types and measure whether full-reference metrics exceed current state-of-the-art agents while inference time stays low and new tools integrate without retraining.

Figures

Figures reproduced from arXiv: 2605.22208 by Jiawei Wu, Kailin Zhuang, Zhi Jin.

Figure 1
Figure 1. Figure 1: Illustration of the experience components in IRA. (Left&Center) We extract embeddings of [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the experience-driven inference workflow in EvoIR-Agent. 3 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the inner closed-loop experience mechanism in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Performance vs. Efficiency trade-off visualiza￾tion [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visual comparison examples on MiO100 dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: (a) Ablation study on experience granularity with three levels; (b) and (c) visualize the [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Performance-Efficiency analysis of the self-evolving process. (a) Quantitative ablation [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visualization of Preference setting. It has been demonstrated that, in instances where image restoration is achieved, the visual quality of the restored image can vary with different tools. Findings on 2. Under degradation coupling setting, jointly exploring the tool T choices and degradation removal order O typically leads to a computationally prohibitive permutational space for experience modeling. We co… view at source ↗
Figure 9
Figure 9. Figure 9: Experiments conducted on the MiO100 dataset and the resulting radar chart. It demonstrates [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Adaptability analysis of EvoIR-Agent variants across out-of-domain (OOD) dataset. The [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: A visualization of the frequency distribution of tool invocation numbers on different [PITH_FULL_IMAGE:figures/full_fig_p018_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Visualization of the BTD Model result in [PITH_FULL_IMAGE:figures/full_fig_p020_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Visualization of the BTD Model result in [PITH_FULL_IMAGE:figures/full_fig_p021_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Performance-Efficiency analysis of the evolution process with batch size [PITH_FULL_IMAGE:figures/full_fig_p022_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: More visual comparison examples on MiO100 dataset. [PITH_FULL_IMAGE:figures/full_fig_p022_15.png] view at source ↗
read the original abstract

Multimodal Large Language Model (MLLM)-driven image restoration agent demonstrates effectiveness in degradation coupling scenarios by flexibly selecting tools and determining removal orders. However, their zero-shot planning often fails without experience, necessitating severe trial-and-error overhead to achieve satisfactory outcomes. Currently, two paradigms are employed to address this issue, yet a dilemma persists: Training-based methods embed intrinsic experience into parameters, achieving high inference efficiency but lacking compatibility with new tools or degradation. In contrast, training-free methods utilize explicit experience storage for compatibility but still incur trial-and-error overhead due to naive experience. To resolve the dilemma, we propose EvoIR-Agent, which first systematically formulates the experience components of a training-free image restoration agent. Subsequently, a hierarchical experience pool is constructed, which enables coarse-to-fine guidance for diverse tools and removal orders. Furthermore, a self-evolving mechanism is introduced to update the pool from scratch using accumulated records, thereby greatly improving performance and efficiency. Extensive experiments reveal that EvoIR-Agent achieves a significant lead in the full reference metrics and yields a remarkable Pareto-optimal balance between performance and efficiency compared to the state-of-the-art methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes EvoIR-Agent, an MLLM-based image restoration agent that formulates explicit experience components, builds a hierarchical experience pool for coarse-to-fine guidance on tool selection and removal orders, and introduces a self-evolving mechanism to update the pool from scratch using accumulated records. This is claimed to resolve the training-based vs. training-free dilemma by improving both performance and efficiency while preserving compatibility, with extensive experiments showing a significant lead in full-reference metrics and a Pareto-optimal performance-efficiency balance over SOTA methods.

Significance. If the self-evolving mechanism and hierarchical pool function as described, the work would provide a concrete path to accumulate and refine agent experience without parameter retraining, addressing a practical limitation in current MLLM-driven restoration agents. The explicit formulation of experience components and the reported Pareto balance are strengths that could influence follow-up agentic systems in low-level vision.

major comments (2)
  1. [§3.3 (Self-evolving mechanism)] The central claim that the self-evolving mechanism 'updates the pool from scratch using accumulated records' and thereby avoids reintroducing trial-and-error overhead rests on the unexamined quality of the first records generated under zero-experience (naive) operation. This initialization step is load-bearing for the claimed resolution of the dilemma, yet the manuscript provides no ablation or iteration-wise analysis showing that early records are sufficiently informative to drive meaningful coarse-to-fine refinement.
  2. [§5.2 (Efficiency analysis)] Table 2 and the associated Pareto-front analysis: the reported efficiency gains are measured after the pool has evolved, but no corresponding measurements are given for the initial zero-experience phase or for the cumulative overhead incurred while the first records are collected. Without these data the efficiency claim cannot be fully evaluated against the training-free baseline.
minor comments (2)
  1. [§3.1] Notation for the hierarchical experience pool (e.g., the distinction between coarse and fine levels) is introduced in §3.1 but not consistently used in the algorithm pseudocode; a single consistent symbol set would improve readability.
  2. [Abstract] The abstract states 'significant lead in the full reference metrics' but does not specify which reference metrics (PSNR, SSIM, LPIPS, etc.) or the exact margins; adding these numbers to the abstract would help readers quickly gauge the improvement.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address the major comments point-by-point below, agreeing to incorporate additional analyses to strengthen the presentation of the self-evolving mechanism and efficiency evaluation.

read point-by-point responses

  1. Referee: [§3.3 (Self-evolving mechanism)] The central claim that the self-evolving mechanism 'updates the pool from scratch using accumulated records' and thereby avoids reintroducing trial-and-error overhead rests on the unexamined quality of the first records generated under zero-experience (naive) operation. This initialization step is load-bearing for the claimed resolution of the dilemma, yet the manuscript provides no ablation or iteration-wise analysis showing that early records are sufficiently informative to drive meaningful coarse-to-fine refinement.

    Authors: We appreciate the referee's point regarding the importance of validating the initialization phase. The mechanism begins with zero-experience and accumulates records from successful restorations to evolve the pool. Although the manuscript emphasizes the final performance after evolution, we recognize that an iteration-wise analysis would better demonstrate how early records enable refinement. We will add this analysis, including performance curves over evolution steps and qualitative examples of record quality, in the revised manuscript. revision: yes

  2. Referee: [§5.2 (Efficiency analysis)] Table 2 and the associated Pareto-front analysis: the reported efficiency gains are measured after the pool has evolved, but no corresponding measurements are given for the initial zero-experience phase or for the cumulative overhead incurred while the first records are collected. Without these data the efficiency claim cannot be fully evaluated against the training-free baseline.

    Authors: We agree that the efficiency analysis should account for the initial phase to provide a complete picture. The reported results focus on the steady-state performance after the pool has evolved, as this is the intended use case. To fully address the concern, we will include additional experiments measuring the overhead during the initial record collection and the cumulative cost compared to training-free methods without evolution. This data will be added to the efficiency section in the revision. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical system design is self-contained

full rationale

The paper describes an architectural proposal for EvoIR-Agent consisting of experience component formulation, a hierarchical experience pool for coarse-to-fine guidance, and a self-evolving update rule driven by accumulated operational records. No mathematical derivations, equations, or parameter-fitting steps are presented that reduce by construction to their own inputs. Performance and efficiency claims rest on experimental results rather than any self-referential definition or self-citation chain. The method is therefore an independent empirical contribution whose validity can be assessed externally via benchmarks, with no load-bearing circularity in the described chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

Based solely on the abstract, the central claim rests on the domain assumption that experience can be explicitly formulated and hierarchically organized to guide agents effectively. The hierarchical experience pool and self-evolving mechanism are the primary new elements introduced.

axioms (1)
  • domain assumption Experience can be systematically formulated into components that enable coarse-to-fine guidance for tools and removal orders in image restoration agents.
    Invoked in the description of the hierarchical experience pool construction.
invented entities (2)
  • Hierarchical experience pool no independent evidence
    purpose: To provide coarse-to-fine guidance for diverse tools and removal orders
    New structure proposed to resolve the dilemma between training-based and training-free methods.
  • Self-evolving mechanism no independent evidence
    purpose: To update the experience pool from accumulated records
    Introduced to improve performance and efficiency over naive experience storage.

pith-pipeline@v0.9.0 · 5730 in / 1373 out tokens · 41800 ms · 2026-05-22T07:45:32.859558+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

92 extracted references · 92 canonical work pages · 10 internal anchors

  1. [1]

    Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising.IEEE transactions on image processing, 26(7): 3142–3155, 2017

    Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang. Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising.IEEE transactions on image processing, 26(7): 3142–3155, 2017

    work page 2017

  2. [2]

    Esrgan: Enhanced super-resolution generative adversarial networks

    Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy. Esrgan: Enhanced super-resolution generative adversarial networks. InProceedings of the European conference on computer vision (ECCV) workshops, pages 0–0, 2018

    work page 2018

  3. [3]

    Scale-recurrent network for deep image deblurring

    Xin Tao, Hongyun Gao, Xiaoyong Shen, Jue Wang, and Jiaya Jia. Scale-recurrent network for deep image deblurring. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 8174–8182, 2018

    work page 2018

  4. [4]

    Mb-taylorformer v2: Improved multi-branch linear transformer expanded by taylor formula for image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    Zhi Jin, Yuwei Qiu, Kaihao Zhang, Hongdong Li, and Wenhan Luo. Mb-taylorformer v2: Improved multi-branch linear transformer expanded by taylor formula for image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    work page 2025

  5. [5]

    Fourllie: Boosting low-light image enhancement by fourier frequency information

    Chenxi Wang, Hongjun Wu, and Zhi Jin. Fourllie: Boosting low-light image enhancement by fourier frequency information. InProceedings of the 31st ACM international conference on multimedia, pages 7459–7469, 2023

    work page 2023

  6. [6]

    A survey on all-in-one image restoration: Taxonomy, evaluation and future trends.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    Junjun Jiang, Zengyuan Zuo, Gang Wu, Kui Jiang, and Xianming Liu. A survey on all-in-one image restoration: Taxonomy, evaluation and future trends.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    work page 2025

  7. [7]

    All in one bad weather removal using architectural search

    Ruoteng Li, Robby T Tan, and Loong-Fah Cheong. All in one bad weather removal using architectural search. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3175–3185, 2020

    work page 2020

  8. [8]

    Transweather: Transformer-based restoration of images degraded by adverse weather conditions

    Jeya Maria Jose Valanarasu, Rajeev Yasarla, and Vishal M Patel. Transweather: Transformer-based restoration of images degraded by adverse weather conditions. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2353–2363, 2022

    work page 2022

  9. [9]

    Instructir: High-quality image restoration following human instructions

    Marcos V Conde, Gregor Geigle, and Radu Timofte. Instructir: High-quality image restoration following human instructions. InEuropean Conference on Computer Vision, pages 1–21. Springer, 2024

    work page 2024

  10. [10]

    Autodir: Automatic all-in-one image restoration with latent diffusion

    Yitong Jiang, Zhaoyang Zhang, Tianfan Xue, and Jinwei Gu. Autodir: Automatic all-in-one image restoration with latent diffusion. InEuropean Conference on Computer Vision, pages 340–359. Springer, 2024

    work page 2024

  11. [11]

    Jarvisir: Elevating autonomous driving perception with intelligent image restoration

    Yunlong Lin, Zixu Lin, Haoyu Chen, Panwang Pan, Chenxin Li, Sixiang Chen, Kairun Wen, Yeying Jin, Wenbo Li, and Xinghao Ding. Jarvisir: Elevating autonomous driving perception with intelligent image restoration. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22369–22380, 2025

    work page 2025

  12. [12]

    Gradient as conditions: Rethinking hog for all-in-one image restoration

    Jiawei Wu, Zhifei Yang, Zhe Wang, and Zhi Jin. Gradient as conditions: Rethinking hog for all-in-one image restoration. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 10682–10690, 2026

    work page 2026

  13. [13]

    Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  14. [14]

    Vision-language models for vision tasks: A survey.IEEE transactions on pattern analysis and machine intelligence, 46(8):5625–5644, 2024

    Jingyi Zhang, Jiaxing Huang, Sheng Jin, and Shijian Lu. Vision-language models for vision tasks: A survey.IEEE transactions on pattern analysis and machine intelligence, 46(8):5625–5644, 2024

    work page 2024

  15. [15]

    Restoreagent: Autonomous image restoration agent via multimodal large language models.Advances in Neural Information Processing Systems, 37:110643–110666, 2024

    Haoyu Chen, Wenbo Li, Jinjin Gu, Jingjing Ren, Sixiang Chen, Tian Ye, Renjing Pei, Kaiwen Zhou, Fenglong Song, and Lei Zhu. Restoreagent: Autonomous image restoration agent via multimodal large language models.Advances in Neural Information Processing Systems, 37:110643–110666, 2024

    work page 2024

  16. [16]

    An intelligent agentic system for complex image restoration problems

    Kaiwen Zhu, Jinjin Gu, Zhiyuan You, Yu Qiao, and Chao Dong. An intelligent agentic system for complex image restoration problems. InThe Thirteenth International Conference on Learning Representations. 10

    work page

  17. [17]

    Restore-R1: Efficient Image Restoration Agents via Reinforcement Learning with Multimodal LLM Perceptual Feedback

    Jianglin Lu, Yuanwei Wu, Ziyi Zhao, Hongcheng Wang, Felix Jimenez, Abrar Majeedi, and Yun Fu. Simplecall: A lightweight image restoration agent in label-free environments with mllm perceptual feedback.arXiv preprint arXiv:2512.18599, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  18. [18]

    Tir-agent: Training an explorative and efficient agent for image restoration

    Yisheng Zhang, Guoli Jia, Haote Hu, Shanxu Zhao, Kaikai Zhao, Long Sun, Xinwei Long, Kai Tian, Che Jiang, Zhaoxiang Liu, et al. Tir-agent: Training an explorative and efficient agent for image restoration. arXiv preprint arXiv:2603.27742, 2026

    work page arXiv 2026

  19. [19]

    Multi-agent image restoration.arXiv preprint arXiv:2503.09403, 2025

    Xu Jiang, Gehui Li, Bin Chen, and Jian Zhang. Multi-agent image restoration.arXiv preprint arXiv:2503.09403, 2025

    work page arXiv 2025

  20. [20]

    4kagent: Agentic any image to 4k super-resolution

    Yushen Zuo, Qi Zheng, Mingyang Wu, Xinrui Jiang, Renjie Li, Jian Wang, Yide Zhang, Gengchen Mai, Lihong Wang, James Zou, et al. 4kagent: Agentic any image to 4k super-resolution. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems

    work page

  21. [21]

    Expel: Llm agents are experiential learners

    Andrew Zhao, Daniel Huang, Quentin Xu, Matthieu Lin, Yong-Jin Liu, and Gao Huang. Expel: Llm agents are experiential learners. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19632–19642, 2024

    work page 2024

  22. [22]

    Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models

    Qizheng Zhang, Changran Hu, Shubhangi Upasani, Boyuan Ma, Fenglu Hong, Vamsidhar Kamanuru, Jay Rainton, Chen Wu, Mengmeng Ji, Hanchen Li, et al. Agentic context engineering: Evolving contexts for self-improving language models.arXiv preprint arXiv:2510.04618, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  23. [23]

    A-MEM: Agentic Memory for LLM Agents

    Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. A-mem: Agentic memory for llm agents.arXiv preprint arXiv:2502.12110, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  24. [24]

    LightMem: Lightweight and Efficient Memory-Augmented Generation

    Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, et al. Lightmem: Lightweight and efficient memory-augmented generation. arXiv preprint arXiv:2510.18866, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  25. [25]

    Llm2clip: Powerful language model unlocks richer cross-modality representation

    Weiquan Huang, Aoqi Wu, Yifan Yang, Xufang Luo, Yuqing Yang, Usman Naseem, Chunyu Wang, Qi Dai, Xiyang Dai, Dongdong Chen, et al. Llm2clip: Powerful language model unlocks richer cross-modality representation. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 5131–5139, 2026

    work page 2026

  26. [26]

    G-memory: Tracing hierarchical memory for multi-agent systems

    Guibin Zhang, Muxin Fu, Kun Wang, Guancheng Wan, Miao Yu, and Shuicheng YAN. G-memory: Tracing hierarchical memory for multi-agent systems. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems

    work page

  27. [27]

    Memweaver: A hierarchical memory from textual interactive behaviors for personalized generation.arXiv preprint arXiv:2510.07713, 2025

    Shuo Yu, Mingyue Cheng, Daoyu Wang, Qi Liu, Zirui Liu, Ze Guo, and Xiaoyu Tao. Memweaver: A hierarchical memory from textual interactive behaviors for personalized generation.arXiv preprint arXiv:2510.07713, 2025

    work page arXiv 2025

  28. [28]

    Pretraining with hierarchical memories: separating long-tail and common knowledge.arXiv preprint arXiv:2510.02375, 2025

    Hadi Pouransari, David Grangier, C Thomas, Michael Kirchhof, and Oncel Tuzel. Pretraining with hierarchical memories: separating long-tail and common knowledge.arXiv preprint arXiv:2510.02375, 2025

    work page arXiv 2025

  29. [29]

    Towards effective multiple-in-one image restoration: A sequential and prompt learning strategy.arXiv preprint arXiv:2401.03379, 2024

    Xiangtao Kong, Chao Dong, and Lei Zhang. Towards effective multiple-in-one image restoration: A sequential and prompt learning strategy.arXiv preprint arXiv:2401.03379, 2024

    work page arXiv 2024

  30. [30]

    Promptir: Prompting for all-in-one image restoration.Advances in neural information processing systems, 36:71275–71293, 2023

    Vaishnav Potlapalli, Syed Waqas Zamir, Salman H Khan, and Fahad Shahbaz Khan. Promptir: Prompting for all-in-one image restoration.Advances in neural information processing systems, 36:71275–71293, 2023

    work page 2023

  31. [31]

    Controlling vision- language models for universal image restoration.arXiv preprint arXiv:2310.01018, 3(8), 2023

    Ziwei Luo, Fredrik K Gustafsson, Zheng Zhao, Jens Sjölund, and Thomas B Schön. Controlling vision- language models for universal image restoration.arXiv preprint arXiv:2310.01018, 3(8), 2023

    work page arXiv 2023

  32. [32]

    Unirestore: Unified perceptual and task-oriented image restoration model using diffusion prior

    I Chen, Wei-Ting Chen, Yu-Wei Liu, Yuan-Chun Chiang, Sy-Yen Kuo, Ming-Hsuan Yang, et al. Unirestore: Unified perceptual and task-oriented image restoration model using diffusion prior. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 17969–17979, 2025

    work page 2025

  33. [33]

    Visual-instructed degradation diffusion for all-in-one image restoration

    Wenyang Luo, Haina Qin, Zewen Chen, Libin Wang, Dandan Zheng, Yuming Li, Yufan Liu, Bing Li, and Weiming Hu. Visual-instructed degradation diffusion for all-in-one image restoration. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 12764–12777, 2025

    work page 2025

  34. [34]

    Learning continuous wasserstein barycenter space for generalized all-in-one image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026

    Xiaole Tang, Xiaoyi He, Jiayi Xu, Xiang Gu, and Jian Sun. Learning continuous wasserstein barycenter space for generalized all-in-one image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026. 11

    work page 2026

  35. [35]

    Q-Agent: Quality-Driven Chain-of-Thought Image Restoration Agent through Robust Multimodal Large Language Model

    Yingjie Zhou, Jiezhang Cao, Zicheng Zhang, Farong Wen, Yanwei Jiang, Jun Jia, Xiaohong Liu, Xiongkuo Min, and Guangtao Zhai. Q-agent: quality-driven chain-of-thought image restoration agent through robust multimodal large language model.arXiv preprint arXiv:2504.07148, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  36. [36]

    Paagent: Portrait-aware image restoration agent via subjective-objective reinforcement learning.arXiv preprint arXiv:2603.17055, 2026

    Yijian Wang, Qingsen Yan, Jiantao Zhou, Duwei Dai, and Wei Dong. Paagent: Portrait-aware image restoration agent via subjective-objective reinforcement learning.arXiv preprint arXiv:2603.17055, 2026

    work page arXiv 2026

  37. [37]

    Hybrid agents for image restoration.arXiv preprint arXiv:2503.10120, 2025

    Bingchen Li, Xin Li, Yiting Lu, and Zhibo Chen. Hybrid agents for image restoration.arXiv preprint arXiv:2503.10120, 2025

    work page arXiv 2025

  38. [38]

    A survey on the memory mechanism of large language model-based agents.ACM Transactions on Information Systems, 43(6):1–47, 2025

    Zeyu Zhang, Quanyu Dai, Xiaohe Bo, Chen Ma, Rui Li, Xu Chen, Jieming Zhu, Zhenhua Dong, and Ji- Rong Wen. A survey on the memory mechanism of large language model-based agents.ACM Transactions on Information Systems, 43(6):1–47, 2025

    work page 2025

  39. [39]

    Rethinking memory mechanisms of foundation agents in the second half.arXiv preprint arXiv:2602.06052, 2026

    Wei-Chieh Huang, Weizhi Zhang, Yueqing Liang, Yuanchen Bei, Yankai Chen, Tao Feng, Xinyu Pan, Zhen Tan, Yu Wang, Tianxin Wei, et al. Rethinking memory mechanisms of foundation agents in the second half.arXiv preprint arXiv:2602.06052, 2026

    work page arXiv 2026

  40. [40]

    How memory management impacts llm agents: An empirical study of experience-following behavior.arXiv preprint arXiv:2505.16067, 2025

    Zidi Xiong, Yuping Lin, Wenya Xie, Pengfei He, Zirui Liu, Jiliang Tang, Himabindu Lakkaraju, and Zhen Xiang. How memory management impacts llm agents: An empirical study of experience-following behavior.arXiv preprint arXiv:2505.16067, 2025

    work page arXiv 2025

  41. [41]

    From experience to strategy: Empowering llm agents with trainable graph memory

    Siyu Xia, Zekun Xu, Jiajun Chai, Wentian Fan, Yan Song, Xiaohan Wang, Guojun Yin, Wei Lin, Haifeng Zhang, and Jun Wang. From experience to strategy: Empowering llm agents with trainable graph memory. arXiv preprint arXiv:2511.07800, 2025

    work page arXiv 2025

  42. [42]

    Skill-Pro: Learning Reusable Skills from Experience via Non-Parametric PPO for LLM Agents

    Qirui Mi, Zhijian Ma, Mengyue Yang, Haoxuan Li, Yisen Wang, Haifeng Zhang, and Jun Wang. Procmem: Learning reusable procedural memory from experience via non-parametric ppo for llm agents.arXiv preprint arXiv:2602.01869, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  43. [43]

    Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolution

    Zouying Cao, Jiaji Deng, Li Yu, Weikang Zhou, Zhaoyang Liu, Bolin Ding, and Hai Zhao. Remember me, refine me: A dynamic procedural memory framework for experience-driven agent evolution.arXiv preprint arXiv:2512.10696, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  44. [44]

    Flex: Continuous agent evolution via forward learning from experience.arXiv preprint arXiv:2511.06449, 2025

    Zhicheng Cai, Xinyuan Guo, Yu Pei, Jiangtao Feng, Jinsong Su, Jiangjie Chen, Ya-Qin Zhang, Wei-Ying Ma, Mingxuan Wang, and Hao Zhou. Flex: Continuous agent evolution via forward learning from experience.arXiv preprint arXiv:2511.06449, 2025

    work page arXiv 2025

  45. [45]

    SimpleMem: Efficient Lifelong Memory for LLM Agents

    Jiaqi Liu, Yaofeng Su, Peng Xia, Siwei Han, Zeyu Zheng, Cihang Xie, Mingyu Ding, and Huaxiu Yao. Simplemem: Efficient lifelong memory for llm agents.arXiv preprint arXiv:2601.02553, 2026

    work page internal anchor Pith review arXiv 2026

  46. [46]

    On extending the bradley-terry model to accommodate ties in paired comparison experiments.Journal of the American Statistical Association, 65(329):317–328, 1970

    Roger R Davidson. On extending the bradley-terry model to accommodate ties in paired comparison experiments.Journal of the American Statistical Association, 65(329):317–328, 1970

    work page 1970

  47. [47]

    Tests of statistical hypotheses concerning several parameters when the number of observations is large.Transactions of the American Mathematical society, 54(3):426–482, 1943

    Abraham Wald. Tests of statistical hypotheses concerning several parameters when the number of observations is large.Transactions of the American Mathematical society, 54(3):426–482, 1943

    work page 1943

  48. [48]

    Chateval: Towards better llm-based evaluators through multi-agent debate

    Chi-Min Chan, Weize Chen, Yusheng Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and Zhiyuan Liu. Chateval: Towards better llm-based evaluators through multi-agent debate. InThe Twelfth International Conference on Learning Representations

    work page

  49. [49]

    Should we be going mad? a look at multi-agent debate strategies for llms

    Andries Petrus Smit, Nathan Grinsztajn, Paul Duckworth, Thomas D Barrett, and Arnu Pretorius. Should we be going mad? a look at multi-agent debate strategies for llms. InInternational Conference on Machine Learning, pages 45883–45905. PMLR, 2024

    work page 2024

  50. [50]

    Bdi agents: from theory to practice

    Anand S Rao, Michael P Georgeff, et al. Bdi agents: from theory to practice. InIcmas, volume 95, pages 312–319, 1995

    work page 1995

  51. [51]

    The proof and measurement of association between two things. 1961

    work page 1961

  52. [52]

    Foundir: Unleashing million-scale training data to advance foundation models for image restoration

    Hao Li, Xiang Chen, Jiangxin Dong, Jinhui Tang, and Jinshan Pan. Foundir: Unleashing million-scale training data to advance foundation models for image restoration. InProceedings of the IEEE/CVF international conference on computer vision, pages 12626–12636, 2025

    work page 2025

  53. [53]

    Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    work page 2004

  54. [54]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 586–595, 2018. 12

    work page 2018

  55. [55]

    Maniqa: Multi-dimension attention network for no-reference image quality assessment

    Sidi Yang, Tianhe Wu, Shuwei Shi, Shanshan Lao, Yuan Gong, Mingdeng Cao, Jiahao Wang, and Yujiu Yang. Maniqa: Multi-dimension attention network for no-reference image quality assessment. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1191–1200, 2022

    work page 2022

  56. [56]

    Exploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Exploring clip for assessing the look and feel of images. InProceedings of the AAAI conference on artificial intelligence, volume 37, pages 2555–2563, 2023

    work page 2023

  57. [57]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. InProceedings of the IEEE/CVF international conference on computer vision, pages 5148–5157, 2021

    work page 2021

  58. [58]

    Depicting beyond scores: Advancing image quality assessment through multi-modal language models

    Zhiyuan You, Zheyuan Li, Jinjin Gu, Zhenfei Yin, Tianfan Xue, and Chao Dong. Depicting beyond scores: Advancing image quality assessment through multi-modal language models. InEuropean Conference on Computer Vision, pages 259–276. Springer, 2024

    work page 2024

  59. [59]

    Q-align: teaching lmms for visual scoring via discrete text-defined levels

    Haoning Wu, Zicheng Zhang, Weixia Zhang, Chaofeng Chen, Liang Liao, Chunyi Li, Yixuan Gao, Annan Wang, Erli Zhang, Wenxiu Sun, et al. Q-align: teaching lmms for visual scoring via discrete text-defined levels. InProceedings of the 41st International Conference on Machine Learning, pages 54015–54029, 2024

    work page 2024

  60. [60]

    Ali M Reza. Realization of the contrast limited adaptive histogram equalization (clahe) for real-time image enhancement.Journal of VLSI signal processing systems for signal, image and video technology, 38(1): 35–44, 2004

    work page 2004

  61. [61]

    Learning to deblur using light field generated and real defocus images

    Lingyan Ruan, Bin Chen, Jizhou Li, and Miuling Lam. Learning to deblur using light field generated and real defocus images. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16304–16313, 2022

    work page 2022

  62. [62]

    Iterative filter adaptive network for single image defocus deblurring

    Junyong Lee, Hyeongseok Son, Jaesung Rim, Sunghyun Cho, and Seungyong Lee. Iterative filter adaptive network for single image defocus deblurring. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2034–2042, 2021

    work page 2034

  63. [63]

    Restormer: Efficient transformer for high-resolution image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5728–5739, 2022

    work page 2022

  64. [64]

    Swinir: Image restoration using swin transformer

    Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. Swinir: Image restoration using swin transformer. InProceedings of the IEEE/CVF international conference on computer vision, pages 1833–1844, 2021

    work page 2021

  65. [65]

    Towards flexible blind jpeg artifacts removal

    Jiaxi Jiang, Kai Zhang, and Radu Timofte. Towards flexible blind jpeg artifacts removal. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4997–5006, 2021

    work page 2021

  66. [66]

    Maxim: Multi-axis mlp for image processing

    Zhengzhong Tu, Hossein Talebi, Han Zhang, Feng Yang, Peyman Milanfar, Alan Bovik, and Yinxiao Li. Maxim: Multi-axis mlp for image processing. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5769–5780, 2022

    work page 2022

  67. [67]

    Multi-stage progressive image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao. Multi-stage progressive image restoration. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14821–14831, 2021

    work page 2021

  68. [68]

    A comparative study of image restoration networks for general backbone network design

    Xiangyu Chen, Zheyuan Li, Yuandong Pu, Yihao Liu, Jiantao Zhou, Yu Qiao, and Chao Dong. A comparative study of image restoration networks for general backbone network design. InEuropean Conference on Computer Vision, pages 74–91. Springer, 2024

    work page 2024

  69. [69]

    Ridcp: Revitalizing real image dehazing via high-quality codebook priors

    Rui-Qi Wu, Zheng-Peng Duan, Chun-Le Guo, Zhi Chai, and Chongyi Li. Ridcp: Revitalizing real image dehazing via high-quality codebook priors. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 22282–22291, 2023

    work page 2023

  70. [70]

    Vision transformers for single image dehazing.IEEE Transactions on Image Processing, 32:1927–1941, 2023

    Yuda Song, Zhuqing He, Hui Qian, and Xin Du. Vision transformers for single image dehazing.IEEE Transactions on Image Processing, 32:1927–1941, 2023

    work page 1927

  71. [71]

    Diffbir: Toward blind image restoration with generative diffusion prior

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Bo Dai, Fanghua Yu, Yu Qiao, Wanli Ouyang, and Chao Dong. Diffbir: Toward blind image restoration with generative diffusion prior. InEuropean conference on computer vision, pages 430–448. Springer, 2024. 13

    work page 2024

  72. [72]

    Hat: Hybrid attention transformer for image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    Xiangyu Chen, Xintao Wang, Wenlong Zhang, Xiangtao Kong, Yu Qiao, Jiantao Zhou, and Chao Dong. Hat: Hybrid attention transformer for image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    work page 2025

  73. [73]

    Diff-plugin: Revitalizing details for diffusion-based low-level tasks

    Yuhao Liu, Zhanghan Ke, Fang Liu, Nanxuan Zhao, and Rynson WH Lau. Diff-plugin: Revitalizing details for diffusion-based low-level tasks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4197–4208, 2024

    work page 2024

  74. [74]

    Fourier priors-guided diffusion for zero-shot joint low-light enhancement and deblurring

    Xiaoqian Lv, Shengping Zhang, Chenyang Wang, Yichen Zheng, Bineng Zhong, Chongyi Li, and Liqiang Nie. Fourier priors-guided diffusion for zero-shot joint low-light enhancement and deblurring. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 25378–25388, 2024

    work page 2024

  75. [75]

    Revitalizing convolutional network for image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12):9423–9438, 2024

    Yuning Cui, Wenqi Ren, Xiaochun Cao, and Alois Knoll. Revitalizing convolutional network for image restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12):9423–9438, 2024

    work page 2024

  76. [76]

    Omni-kernel network for image restoration

    Yuning Cui, Wenqi Ren, and Alois Knoll. Omni-kernel network for image restoration. InProceedings of the AAAI conference on artificial intelligence, volume 38, pages 1426–1434, 2024

    work page 2024

  77. [77]

    Simple baselines for image restoration

    Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. In European conference on computer vision, pages 17–33. Springer, 2022

    work page 2022

  78. [78]

    Efficient visual state space model for image deblurring

    Lingshun Kong, Jiangxin Dong, Jinhui Tang, Ming-Hsuan Yang, and Jinshan Pan. Efficient visual state space model for image deblurring. InProceedings of the computer vision and pattern recognition conference, pages 12710–12719, 2025

    work page 2025

  79. [79]

    Drct: Saving image super-resolution away from information bottleneck

    Chih-Chung Hsu, Chia-Ming Lee, and Yi-Shiuan Chou. Drct: Saving image super-resolution away from information bottleneck. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6133–6142, 2024

    work page 2024

  80. [80]

    Hmanet: Hybrid multi-axis aggregation network for image super-resolution

    Shu-Chuan Chu, Zhi-Chao Dou, Jeng-Shyang Pan, Shaowei Weng, and Junbao Li. Hmanet: Hybrid multi-axis aggregation network for image super-resolution. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6257–6266, 2024

    work page 2024

Showing first 80 references.