Pith. sign in

REVIEW 3 major objections 4 minor 48 references

Flow-Map Distillation on Relation Manifolds for Image Restoration

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read FoRM turns image-restoration distillation into a flow-mapping problem on relation manifolds.

desk verdict Solid empirical KD work with honest ablations, but the flow-map operator is underdetermined and the 'semigroup' loss is misnamed; the paper needs a reframing before the mechanism can be trusted. read the letter →

arxiv 2608.05769 v1 pith:GTGOG24T submitted 2026-08-06 cs.CV

Zihao He , Songhua Liu This is my paper · ORCID
classification cs.CV
keywords knowledgedistillationimagerestorationflowmatchingrelationmanifoldsemigroupconsistencytrainingstabilitysuper-resolutionmapoperator
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

FoRM claims that knowledge distillation for image restoration should be modelled as a continuous flow-mapping problem rather than as static endpoint matching. It learns an operator F_theta(z,t,s) that predicts the teacher's relation state at any time s from the student's relation state at time t, along the linear interpolation bridge between them. Two structural constraints, safe semigroup consistency and endpoint anchoring, keep this operator compositionally self-consistent and anchored to the teacher target. Across five restoration tasks and several backbone architectures, the paper reports consistent quality gains over existing distillation baselines and roughly halved late-training PSNR variance compared with naive flow-matching distillation. If correct, this makes trajectory-level supervision a practical, inference-free-cost addition to image-restoration pipelines.

What carries the argument

The central object is the flow map operator F_theta(z,t,s), a three-layer MLP with SiLU activations that maps a relation state z at source time t to the predicted relation state at target time s, using independent sinusoidal time embeddings and a learned layer-identity embedding. It is trained by three losses: dense map distillation L_map on arbitrary (t,s) pairs, endpoint anchoring L_bd at s=1, and safe semigroup consistency L_sg that enforces F_theta(z_t,t,1) approximately equals F_theta(z_s,s,1) using ground-truth bridge points and stop-gradient. The relation manifold is the product of simplices ($\Delta$^(N-1))^N, where each row of the softmax affinity matrix lies on a probability simplex, so linear interpolation between relation states remains a valid relation configuration.

What would settle it

On a trained model, collect many (z_t, z_s) pairs for fixed (t,s) across images; if the same z_t is paired with materially different teacher states z_s, then F_theta cannot be a well-defined function of z_t alone and the central mechanism loses its justification. A simpler check is to ablate the safe semigroup loss while keeping the map and anchoring losses, and see whether the PSNR and late-epoch variance gains persist on all five tasks; if they do, the semigroup constraint is not load-bearing.

Watch

Extended reading notes

Core claim

FoRM's central claim is that the student-teacher knowledge transfer should be supervised along the whole trajectory of relation states, not just at the endpoint. The distillation state is the vectorized softmax-normalized pairwise token-affinity matrix of pooled intermediate features; student and teacher states are the endpoints of a line bridge that stays inside the product-of-simplices relation manifold. A shared three-layer MLP, conditioned on sinusoidal time embeddings and a layer-identity embedding, learns F_theta(z,t,s) to predict the relation state at any target time s. The paper argues that the semigroup property, enforced safely by conditioning both branches on ground-truth bridge points with a stop-gradient, prevents the phantom-state error that recursive self-distillation suffers, and that endpoint anchoring at s=1 prevents drift from the teacher target. Experiments across super-resolution, deraining, denoising, deblurring, and low-light enhancement show consistent gains over state-of-the-art distillation baselines, with the flow-map operator discarded at inference time.

Load-bearing premise

The flow-map operator assumes that a student relation state at time t is enough to determine the teacher relation state at any later time s, but the paper does not prove that the teacher state or velocity is recoverable from a single intermediate relation state.

Editorial extensions

If this is right

  • FoRM improves PSNR/SSIM over static relation distillation and over constant-velocity flow matching on all five restoration tasks tested.
  • Late-training PSNR variance drops by about half relative to naive flow-matching distillation, and the deployed student costs nothing extra because F_theta is discarded after training.
  • The method transfers across backbone architectures including RCAN, SwinIR, EDSR, and a cross-architecture teacher-student pair, so it acts as a plug-in distillation module.
  • Component ablations indicate that endpoint anchoring is a precondition for the semigroup loss; applying semigroup consistency without anchoring worsens training stability, so the two structural constraints are mutually dependent.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension is to apply FoRM's safe semigroup idea to other iterative distillation and self-distillation settings where predicted states are recursively fed back, since the phantom-state error mechanism is not specific to image restoration.
  • The identifiability question, whether the student relation state at time t alone determines the teacher state at time s, is not settled by the paper; if it fails, the operator learns an average over disagreeing targets, which a variance-sensitive diagnostic could reveal.
  • The relation-manifold formulation suggests that pooling size p and layer set L tune a trade-off between supervision richness and memory, and p=8 may not be optimal for datasets with stronger long-range structure or higher-resolution relations.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes FoRM, a knowledge-distillation method for image restoration in which an auxiliary flow-map operator F_theta(z,t,s) is trained to predict relation states at arbitrary times along a linear interpolation bridge between student and teacher relation matrices. Three losses are used: a map-distillation loss, an endpoint-anchoring loss, and a 'safe semigroup consistency' loss that evaluates the flow map on ground-truth bridge points. The method is evaluated on five restoration tasks with several backbone architectures, reporting small but consistent PSNR/SSIM gains over prior distillation baselines and reduced late-training variance. The central theoretical claims are that F_theta defines a well-posed flow map and that Eq. (6) enforces the semigroup property of Eq. (5).

Significance. If the flow-map formulation were well posed, the paper would offer a genuinely new perspective on trajectory-level knowledge distillation for image restoration, and the experimental breadth—five tasks, multiple backbones, explicit stability metrics, and component ablations—is a clear strength. The stability comparison with naive flow matching and the safe-versus-naive semigroup ablation are informative and well designed. However, the central operator F_theta is not a well-defined function on its stated input, and the proposed 'safe semigroup consistency' loss does not enforce the semigroup property it claims. These issues undermine the theoretical foundation of the method rather than being local presentation problems, so the contribution as it stands is not established.

major comments (3)
  1. [Sec. 3.2, Eqs. (3)-(4)] The flow-map operator F_theta is not a well-defined function on its stated input. The target is z_s = (1-s)z_0 + s z_1, while the input is h_l = [z_t || e_t || e_s || e_l] with z_t = (1-t)z_0 + t z_1. For fixed t and s, z_s is not determined by (z_t,t,s): at t=0 the input is exactly z_0 and the target still depends on z_1; for t>0, z_s = (s/t) z_t + ((1-s) - s(1-t)/t) z_0, which still depends on z_0. Since z_0 and z_1 vary across images and training steps, two examples can share the same operator input but require different targets. Consequently L_map in Eq. (7) is fitting a multi-valued relation, not a function, and no identifiability argument or inclusion of the missing endpoint in the input is provided. This is load-bearing because the paper's central claim is that F_theta directly predicts the relation state at any target time from the current state alone.
  2. [Sec. 3.3, Eq. (6)] The 'safe semigroup consistency' loss does not implement the semigroup property stated in Eq. (5). Eq. (5) requires F(z_t,t,1) = F(F(z_t,t,s),s,1), i.e., the outer map must be evaluated at the predicted intermediate state F(z_t,t,s). Eq. (6) instead compares F(z_t,t,1) with F(z_s,s,1) using the ground-truth bridge point z_s. This is a self-consistency constraint along the true linear bridge, not a compositional semigroup constraint; it never evaluates the operator on its own output, so it cannot enforce compositional agreement. The claim that the loss 'preserves the compositional structure' of the semigroup property is therefore not supported by the stated objective.
  3. [Tables 2 and 3] The empirical claim of consistent gains is not backed by a statistical comparison. In Table 2, many FoRM advantages over the strongest baseline DCKD are 0.02-0.09 dB on individual benchmarks, while Table 3 reports a cross-seed standard deviation of 0.0112 dB for FoRM on Set5. No confidence intervals or significance tests are reported for the PSNR/SSIM comparisons in Tables 1 and 2, so several of the reported margins are comparable to or smaller than the seed noise. The CKA analysis in Fig. 7 does use a Wilcoxon test, but the primary reconstruction-quality claims do not.
minor comments (4)
  1. [Abstract and Sec. 4.3, Table 3] The abstract states that FoRM reduces training variance by 'approximately 50%' compared to naive flow matching distillation, but Table 3 shows a 78% reduction in sigma_late relative to Naive FM and a 52% reduction relative to Static Rel. KD; the text in Sec. 4.3 correctly reports 78%, so the abstract should be corrected.
  2. [Table 4] Configurations B0, B1, and A0 all have identical entries in the L_map, L_bd, and L_sg columns, so the table does not convey the differences among raw-feature KD, static relation KD, and naive flow matching; the caption or table should encode the representation and objective type explicitly.
  3. [Sec. 3.2] The paper repeatedly calls the state space the 'relation manifold' but never constrains F_theta's output to lie on (Delta^{N-1})^N; since F_theta is a plain MLP, its predictions are generally not valid relation states, so the manifold interpretation applies only to the targets z_0, z_1, z_t, z_s, not to the operator's outputs.
  4. [Table 1] The column labeled 'Average Score' appears to be the average PSNR and average SSIM separately, but the caption does not say this; please clarify the averaging convention.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the flow-map supervision is a learned regression on constructed interpolation targets, not a reduction to inputs; flagged well-posedness issues are correctness concerns, not circularity.

full rationale

The paper's central training objective is a regression of z_s (linear interpolation of student and teacher relation states) from an operator input that contains z_t and time/layer embeddings. This is not circular: the target z_s is not implied by the input h_l by construction, and the MLP must learn the mapping. The fact that the input omits the teacher endpoint z_1 (and generally cannot identify z_0,z_1 from z_t) is an identifiability/well-posedness concern, not a circular reduction, because the loss is not automatically minimized by the input. The 'safe semigroup' loss in Eq. (6) compares two direct predictions F(z_t,t,1) and F(z_s,s,1) using a ground-truth bridge point, and it does not literally implement the compositional semigroup Eq. (5); this is a correctness or labeling gap rather than a case of the output being equal to the input by definition. Endpoint anchoring in Eq. (7) is straightforward supervised regression to the teacher relation state. Self-citations in the paper ([11], [36], [37]) are confined to experimental protocol and related generative restoration work, while the flow-map and semigroup ideas are credited to external works [4,29,30]; no load-bearing self-citation chain is present. Final claims are evaluated on standard restoration benchmarks (PSNR/SSIM) with ablations, so the empirical content is independent of the constructed bridge. No step in the derivation reduces to its own inputs; score 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method's central construction rests on several choices that are either standard assumptions or ad hoc to the paper. The free parameters are hyperparameters chosen by hand or by small ablations; none are fitted to a physical law. The key ad hoc assumption is that conditioning the consistency loss on ground-truth linear interpolation points preserves the semigroup structure, which is not demonstrated. No new physical entities are introduced.

free parameters (5)
  • Softmax temperature tau = 0.07
    Fixed following MoCo convention; controls the sharpness of relation affinity matrices and therefore all state distances used in the losses.
  • Loss weights lambda_map, lambda_bd, lambda_sg = 1.0, 0.5, 0.1
    Chosen manually; no sensitivity analysis is reported in the paper.
  • Schedule constants rho0 and w_min = 0.6 and 0.1
    Decay onset and minimum weight in the KD schedule; justified only by an on/off ablation, not by a sweep.
  • Source time sampling range = U[0, 0.8]
    Selected to avoid degenerate consistency intervals near t=1; ablation shows +0.09 dB over U[0,1].
  • Relation pooling size p = 8
    Ablated; p=8 gives the best PSNR while memory grows as O(p^4).
assumptions (5)
  • standard math The product of simplices (Delta^{N-1})^N is convex, so linear interpolation between student and teacher relation states stays within the relation manifold.
    Used in Sec. 3.2 to justify ground-truth bridge states; mathematically correct.
  • domain assumption A single flow-map MLP with layer embeddings can represent the mapping from relation states at all selected layers.
    Assumed in Sec. 3.2; the ablation only compares shared vs per-layer heads, not whether the mapping is well-defined.
  • ad hoc to paper Ground-truth bridge states provide a safe substitute that preserves the compositional structure of the semigroup property.
    This is the load-bearing assumption behind Lsg; the paper does not prove that replacing predicted states with GT linear-interpolation points enforces Eq. (5) on the learned operator.
  • domain assumption Relation-matrix alignment is a sufficient proxy for teacher knowledge in image restoration KD.
    Adopted from relation KD literature; plausible but not established in this paper.
  • domain assumption The teacher is frozen while student, projectors, and map head are trained.
    Standard in KD; not an issue.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Flow-Map Distillation on Relation Manifolds for Image Restoration." pith.science (2026). https://pith.science/paper/GTGOG24T

@misc{pith2026260805769,
  author       = {Pith},
  title        = {Pith review of: Flow-Map Distillation on Relation Manifolds for Image Restoration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GTGOG24T}},
  note         = {Machine review of arXiv:2608.05769}
}
abstract

Knowledge distillation for image restoration typically aligns intermediate features or relation matrices between teacher and student networks as static targets, ignoring the dynamic structure of the knowledge transfer process. In this paper, we propose Flow-Map Distillation on Relation Manifolds (FoRM), which reformulates relation-based knowledge transfer as a continuous flow mapping problem on the relation manifold. Rather than regressing a constant velocity field between student and teacher relation states, FoRM learns a flow map operator $\mathcal{F}_\theta(\mathbf{z}, t, s)$ that directly predicts the relation state at any target time $s$ given the current state at time $t$, enabling richer trajectory-level supervision. To ensure global self-consistency of the learned flow map, we introduce a safe semigroup consistency constraint that enforces compositional agreement using ground-truth bridge states, eliminating phantom-state error accumulation. An endpoint anchoring loss further prevents the operator from drifting away from the teacher target. Extensive experiments on five image restoration tasks, including super-resolution, deraining, denoising, deblurring, and low-light enhancement, demonstrate consistent gains over state-of-the-art distillation baselines across multiple backbone architectures, reducing training variance by approximately 50\% compared to naive flow matching distillation while achieving superior restoration quality.

Figures

Figures reproduced from arXiv: 2608.05769 by the authors.

Figure 1
Figure 1. Motivation and overview of FoRM. Left: Static KD forces the student to directly match the teacher without in￾termediate guidance, leaving the optimisation path unde￾fined. Middle: FoRM constructs structured flow-map anchors z1, z2, z3 along the distillation trajectory. Right: Relation map comparison on Urban100. FoRM achieves significantly higher top-6 patch overlap (71% vs. 33%) and CKA (0.9847 vs. 0.8153) with the… view at source ↗
Figure 2
Figure 2. Overview of the FoRM framework. Teacher and student features are extracted at layers [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison on four image restoration tasks. For deraining ( [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on composite image degradation (CDD-11 [ [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison on low-light enhancement [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison on Urban100 (×4). For each example, we show the HR patch (ground truth), super-resolved patches from the Teacher, Static KD, and FoRM (Ours), along with their corresponding relation maps. FoRM achieves higher CKA with the Teacher’s relation map (…
Figure 7
Figure 7. Figure 7: Per-image CKA with Teacher across three layers [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 39 canonical work pages

  1. [1]

    Eirikur Agustsson and Radu Timofte. 2017. NTIRE 2017 challenge on single image super-resolution: Dataset and study. InCVPRW

    work page 2017

  2. [2]

    Michael S Albergo and Eric Vanden-Eijnden. 2023. Building normalizing flows with stochastic interpolants. InICLR

    work page 2023

  3. [3]

    Marco Bevilacqua, Aline Roumy, Christine Guillemot, and Marie-Line Alberi- Morel. 2012. Low-complexity single-image super-resolution based on nonnega- tive neighbor embedding. InBMVC

    work page 2012

  4. [4]

    Nicholas M Boffi, Michael S Albergo, and Eric Vanden-Eijnden. 2025. How to build a consistency model: Learning flow maps via self-distillation. InNeurIPS

    work page 2025

  5. [5]

    Pengguang Chen, Shu Liu, Hengshuang Zhao, and Jiaya Jia. 2021. Distilling knowledge via knowledge review. InCVPR

    work page 2021

  6. [6]

    Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. 2014. Learning a deep convolutional network for image super-resolution. InECCV

    work page 2014

  7. [7]

    Jianping Gou, Baosheng Yu, Stephen J Maybank, and Dacheng Tao. 2021. Knowl- edge distillation: A survey.International journal of computer vision129, 6 (2021), 1789–1819

    2021

  8. [8]

    Yu Guo, Yuan Gao, Yuxu Lu, Huilin Zhu, Ryan Wen Liu, and Shengfeng He. 2024. OneRestore: A Universal Restoration Framework for Composite Degradation. In European Conference on Computer Vision (ECCV). 255–272. doi:10.1007/978-3- 031-72655-2_15

    doi:10.1007/978-3- 2024

Show all 48 references
  1. [9]

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. 2020. Mo- mentum contrast for unsupervised visual representation learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 9729–9738

    2020

  2. [10]

    Zibin He, Tao Dai, Jian Lu, Yong Jiang, and Shu-Tao Xia. 2020. FAKD: Feature- affinity based knowledge distillation for efficient image super-resolution. InICIP. 518–522

    2020

  3. [11]

    Zihao He, Yunfeng Wu, Xinchao Wang, and Songhua Liu. 2026. Bend the Basics: Degradation-Aware Deformable Tokenization for All-in-One Image Restoration. InICML

    2026

  4. [12]

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531(2015)

    2015 arXiv

  5. [13]

    Jia-Bin Huang, Abhishek Singh, and Narendra Ahuja. 2015. Single image super- resolution from transformed self-exemplars. InCVPR

    2015

  6. [14]

    Yuxuan Jiang, Chen Feng, Fan Zhang, and David Bull. 2024. Mtkd: Multi-teacher knowledge distillation for image super-resolution. InEuropean Conference on Computer Vision. Springer, 364–382

    2024

  7. [15]

    Dongjun Kim, Chieh-Hsin Lai, Wei-Hsiang Liao, Naoki Murata, Yuhta Takida, Toshimitsu Uesaka, Yutong He, Yuki Mitsufuji, and Stefano Ermon. 2024. Consis- tency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion. InICLR

    2024

  8. [16]

    Diederik P Kingma and Jimmy Ba. 2015. Adam: A method for stochastic opti- mization. InICLR

    2015

  9. [17]

    Boyi Li, Wenqi Ren, Dengpan Fu, Dacheng Tao, Dan Feng, Wenjun Zeng, and Zhangyang Wang. 2019. Benchmarking Single-Image Dehazing and Beyond. IEEE Transactions on Image Processing28, 1 (2019), 492–505

    2019

  10. [18]

    Jianze Li, Jiezhang Cao, Yong Guo, Wenbo Li, and Yulun Zhang. 2025. One Diffusion Step to Real-World Super-Resolution via Flow Trajectory Distillation. InICML

    2025

  11. [19]

    Simiao Li, Yun Zhang, Wei Li, Hanting Chen, Wenjia Wang, Bingyi Jing, Shaohui Lin, and Jie Hu. 2025. Knowledge distillation with multi-granularity mixture of priors for image super-resolution. InICLR

    2025

  12. [20]

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

    2021

  13. [21]

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. 2017. Enhanced deep residual networks for single image super-resolution. InCVPRW. 136–144

    2017

  14. [22]

    Yaron Lipman, Ricky T. Q. Chen, Heli Ben-Hamu, Maximilian Nickel, and Matthew Le. 2023. Flow matching for generative modeling. InICLR

    2023

  15. [23]

    Xingchao Liu, Chengyue Gong, and Qiang Liu. 2023. Flow straight and fast: Learning to generate and transfer data with rectified flow. InICLR

    2023

  16. [24]

    Amir M Mansourian, Rozhan Ahmadi, Masoud Ghafouri, Amir Mohammad Babaei, Elaheh Badali Golezani, Zeynab Yasamani Ghamchi, Vida Ramezanian, Alireza Taherian, Kimia Dinashi, Amirali Miri, and Shohreh Kasaei. 2025. A com- prehensive survey on knowledge distillation.Transactions o...

    2025

  17. [25]

    David Martin, Charless Fowlkes, Doron Tal, and Jitendra Malik. 2001. A database of human segmented natural images and its application to evaluating segmenta- tion algorithms and measuring ecological statistics. InICCV, Vol. 2. 416–423

    2001

  18. [26]

    Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. 2017. Deep Multi-Scale Convolutional Neural Network for Dynamic Scene Deblurring. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    2017

  19. [27]

    Wonpyo Park, Dongju Kim, Yan Lu, and Minsu Cho. 2019. Relational knowledge distillation. InCVPR

    2019

  20. [28]

    Adriana Romero, Nicolas Ballas, Samira Ebrahimi Kahou, Antoine Chassang, Carlo Gatta, and Yoshua Bengio. 2015. Fitnets: Hints for thin deep nets. InICLR

    2015

  21. [29]

    Amirmojtaba Sabour, Sanja Fidler, and Karsten Kreis. 2025. Align Your Flow: Scaling Continuous-Time Flow Map Distillation. InNeurIPS

    2025

  22. [30]

    Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. 2023. Consistency models. InICML

    2023

  23. [31]

    Yonglong Tian, Dilip Krishnan, and Phillip Isola. 2020. Contrastive representation distillation. InICLR

    2020

  24. [32]

    Frederick Tung and Greg Mori. 2019. Similarity-preserving knowledge distillation. InICCV

    2019

  25. [33]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need.Advances in neural information processing systems30 (2017)

    2017

  26. [34]

    Yanbo Wang, Shaohui Lin, Yanyun Qu, Haiyan Wu, Zhizhong Zhang, Yuan Xie, and Angela Yao. 2021. Towards compact single image super-resolution via contrastive self-distillation. InIJCAI. 1122–1128

    2021

  27. [35]

    Chen Wei, Wenjing Wang, Wenhan Yang, and Jiaying Liu. 2018. Deep Retinex Decomposition for Low-Light Enhancement. InBritish Machine Vision Conference (BMVC)

    2018

  28. [36]

    Yunfeng Wu, Hongying Cheng, Zihao He, and Songhua Liu. 2026. ViBe: Ultra-High-Resolution Video Synthesis Born from Pure Images.arXiv preprint arXiv:2603.23326(2026)

    2026

  29. [37]

    Yunfeng Wu, Jiayi Song, Zhenxiong Tan, Zihao He, and Songhua Liu. 2025. FreeSwim: Revisiting Sliding-Window Attention Mechanisms for Training-Free Ultra-High-Resolution Video Generation.arXiv preprint arXiv:2511.14712(2025)

    2025

  30. [38]

    Jiaqi Xu, Wenbo Li, Haoze Sun, Fan Li, Zhixin Wang, Long Peng, Jingjing Ren, Haoran Yang, Xiaowei Hu, Renjing Pei, and Pheng-Ann Heng. 2025. Fast Image Super-Resolution via Consistency Rectified Flow. InICCV. 11755–11765

    2025

  31. [39]

    Donglin Yang, Yongxing Zhang, Xin Yu, Liang Hou, Xin Tao, Pengfei Wan, Xiaojuan Qi, and Renjie Liao. 2026. Stable Velocity: A Variance Perspective on Flow Matching. InICML

    2026

  32. [40]

    Tan, Jiashi Feng, Jiaying Liu, Zongming Guo, and Shuicheng Yan

    Wenhan Yang, Robby T. Tan, Jiashi Feng, Jiaying Liu, Zongming Guo, and Shuicheng Yan. 2017. Deep Joint Rain Detection and Removal From a Single Image. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    2017

  33. [41]

    Weiyi You, Mingyang Zhang, Leheng Zhang, Xingyu Zhou, Kexuan Shi, and Shuhang Gu. 2025. Consistency Trajectory Matching for One-Step Generative Super-Resolution. InICCV. 12747–12756

    2025

  34. [42]

    Sergey Zagoruyko and Nikos Komodakis. 2017. Paying more attention to atten- tion: Improving the performance of convolutional neural networks via attention transfer. InICLR

    2017

  35. [43]

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. 2022. Restormer: Efficient Transformer for High- Resolution Image Restoration. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 5728–5739

    2022

  36. [44]

    Roman Zeyde, Michael Elad, and Matan Protter. 2010. On single image scale-up using sparse-representations. InCurves and Surfaces

    2010

  37. [45]

    Weijia Zhang, Fei Xie, Weidong Cai, and Chao Ma. 2025. VRM: Knowledge Distillation via Virtual Relation Matching. InICCV. 2707–2717

    2025

  38. [46]

    Yulun Zhang, Kunpeng Li, Kai Li, Lichen Wang, Bineng Zhong, and Yun Fu. 2018. Image super-resolution using very deep residual channel attention networks. In ECCV. 294–310

    2018

  39. [47]

    Yun Zhang, Wei Li, Simiao Li, Hanting Chen, Zhijun Tu, Wenjia Wang, Bingyi Jing, Shaohui Lin, and Jie Hu. 2023. Data upcycling knowledge distillation for image super-resolution.arXiv preprint arXiv:2309.14162(2023)

    2023 arXiv

  40. [48]

    Yunshuai Zhou, Junbo Qiao, Jincheng Liao, Wei Li, Simiao Li, Jiao Xie, Yunhang Shen, Jie Hu, and Shaohui Lin. 2025. Dynamic contrastive knowledge distillation for efficient image restoration. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 10861–10869

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.