Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Uniformity First: Uniformity-aware Test-time Adaptation of Vision-language Models against Image Corruption

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read CLIP's image embeddings lose uniformity — a marker of lost information — when images are corrupted by blur, noise, or weather, and the paper shows test-time adaptation must update the image encoder itself, not just prompts or embeddings.

desk verdict Useful empirical result with a broken theoretical motivation: the uniformity-based TTA works modestly, but the Section III information story is wrong and hyperparameters leak the test distribution. read the letter →

arxiv 2505.12912 v1 pith:S2FED67D submitted 2025-05-19 cs.CV

Kazuki Adachi , Shin'ya Yamaguchi , Tomoki Hamagami This is my paper · ORCID
classification cs.CV
keywords test-timeadaptationvision-languagemodelsCLIPimagecorruptionsensordegradationuniformitylow-rankzero-shotclassification
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

The paper sets out to show that CLIP's zero-shot classification fails on sensor-corrupted images (blur, noise, weather, compression artifacts) for a reason earlier test-time adaptation methods do not address: the image encoder's embeddings lose uniformity, which the paper reads as a loss of retained input information, whereas under ordinary domain shifts the embeddings stay informative and only the modality gap between image and text matters. It matters because deployed systems such as autonomous driving and surveillance face unexpected sensor degradation, and existing test-time fixes that tune prompts or embeddings leave the encoder fixed and so cannot restore information that was never encoded. The paper's method, UnInfo, updates the image encoder with low-rank adapters under an information-balanced loss that restores uniformity first and sharpens prediction confidence afterward, with an EMA teacher for stability. On ImageNet-C and the deliberately dissimilar ImageNet-C-bar, UnInfo achieves the best mean accuracy among the compared methods (27.10% and 36.13%), with the largest gains on exactly the corruption types where uniformity loss is highest. If the claim holds, the practical lesson is that test-time adaptation to corrupted imagery should act on the encoder itself and treat information retention, not prompt alignment, as the primary target.

What carries the argument

The load-bearing object is the uniformity loss $L_{\mathrm{unif}} = \log \frac{1}{B^2} \sum_{i,j} \exp(-\|z_i - z_j\|_2^2)$, which measures how evenly the batch's image embeddings spread over the unit hypersphere; a lower value is read as more retained input information, an interpretation the paper supports with the data-processing-inequality bound $H(Z) \le H(X) - H(X \mid X')$. Around this it builds the information-balanced objective $\min_{\phi_{\mathrm{img}}} w L_{\mathrm{ent}} + \lambda w^{-1} L_{\mathrm{unif}} + L_{\mathrm{pl}}$, in which the weight $w = \exp(I(z;\hat{y}) - I_0)$ uses the mutual information between embeddings and predictions to decide whether to push uniformity first (severe corruption) or confidence first (mild corruption). The image encoder is updated only through LoRA adapters on the attention layers, and an EMA teacher supplies both the distillation target $L_{\mathrm{pl}}$ and the weights used at inference, preventing the collapse to a single-class solution that pure entropy minimization produces.

What would settle it

Compute a ground-truth information measure for the same corruptions — for instance, the mutual information between the image embeddings and the true class labels, or the accuracy of a fixed linear probe trained on clean ImageNet features — and track it alongside the uniformity loss and UnInfo's accuracy across all 15 ImageNet-C corruptions at several severities; the central claim predicts that improvements in uniformity coincide with recoveries of label-relevant information, so a corruption where uniformity rises without the probe recovering, or where accuracy improves without uniformity improving, would falsify the mechanism.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that sensor degradation damages CLIP differently from ordinary domain shifts: it does not much widen the modality gap between image and text embeddings, but it does corrupt the image embeddings in terms of uniformity — the embeddings clump instead of spreading over the unit hypersphere, and the uniformity loss rises from 0.513 on clean ImageNet to 0.855 under Gaussian noise while prediction entropy climbs. The authors read this through the data processing inequality: for the corruption chain $X \to X' \to Z$, the mutual information bound implies $H(Z) \le H(X) - H(X \mid X')$, so corruption shrinks the upper bound on the entropy of the embedding, meaning less input information is retained. Because previously proposed TTA methods for CLIP leave the image encoder fixed and only modify prompts, text embeddings, or predictions, they cannot recover information that was never encoded; even appending corruption names to the prompt fails because CLIP cannot reliably recognize corruption types. UnInfo instead updates the image encoder with LoRA, minimizing an information-balanced objective $w L_{\mathrm{ent}} + \lambda w^{-1} L_{\mathrm{unif}} + L_{\mathrm{pl}}$ whose weight $w = \exp(I(z;\hat{y}) - I_0)$ is set by the mutual information between embeddings and predictions, so that uniformity is restored first under severe corruption and confidence is sharpened afterward. The same objective, stabilized by an EMA teacher used both for distillation and for inference, attains the highest mean accuracy among all compared methods on ImageNet-C (27.10%) and on ImageNet-C-bar (36.13%).

Load-bearing premise

The entirety of the method rests on treating the uniformity loss — how evenly the adapted image embeddings spread over the unit sphere — as a faithful measure of how much classification-relevant information the encoder has retained; if that proxy separates from the true information content of the embeddings, the information-balancing argument loses its grounding even though the accuracy gains might persist.

Editorial extensions

If this is right

  • Prompt-side fixes are almost ineffective under sensor degradation: ensembling 80 prompts gains about one percentage point, and adding corruption names to the prompt can lower accuracy; only updating the image encoder recovers the lost information.
  • The effective order of operations depends on corruption severity: uniformity must be restored before entropy is minimized for severe corruption such as Gaussian noise, while entropy can be minimized first for mild corruption such as brightness, and the mutual-information weight $w = \exp(I(z;\hat{y}) - I_0)$ supplies this ordering without knowing the corruption type.
  • Entropy minimization alone is destructive on corrupted inputs — the ablation collapses to 0.10% accuracy, effectively predicting a single class — so practical test-time adaptation under corruption needs a stabilizing mechanism such as the EMA-teacher distillation.
  • Because the adapted knowledge accumulates in the LoRA parameters, adaptation can be stopped once the distribution is stable and the LoRA merged into the base model, making inference as cheap as no adaptation at all.
  • The accuracy gains concentrate on the corruption types with the highest uniformity loss (blur and noise), consistent with the claim that the mechanism is information retention rather than generic feature tuning.

Reading between the lines

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

  • A testable extension the paper leaves implicit: the mutual-information weight $w = \exp(I(z;\hat{y}) - I_0)$ could double as an unsupervised domain-change detector — a sudden drop in $I(z;\hat{y})$ would signal the onset of a new corruption type and could trigger re-weighting in the continual or fully online setting the authors name as future work.
  • Because the paper links uniformity to information through an inequality rather than an equality, a calibration study that measures the mutual information between adapted embeddings and clean class labels as the uniformity loss changes would turn the proxy into a quantitative law that later methods could optimize directly.
  • If uniformity is the real bottleneck, the same loss-balancing recipe should transfer to other encoders and to mixed or multi-severity corruptions, where prompt-based methods have even less leverage and the predictability of the method ranking would give a sharp test of the theory.
  • The results on ImageNet-C-bar, which was designed to be dissimilar from ImageNet-C, suggest the mechanism is not overfitting to a corruption family; pushing the same procedure to adversarial or other out-of-distribution inputs that also collapse uniformity would show whether the improvement is specific to sensor degradation.
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 / 6 minor

Summary. The paper studies test-time adaptation (TTA) of CLIP for zero-shot classification under image corruption, which the authors call sensor degradation. Through a preliminary experiment on ImageNet-C, they observe that corrupted images increase both prediction entropy and a uniformity loss over image embeddings, and that simple prompt ensembling does not recover accuracy. They propose UnInfo, which updates the image encoder with LoRA adapters using an objective that combines entropy minimization, a uniformity loss, and knowledge distillation from an EMA teacher. The entropy and uniformity terms are balanced by an adaptive weight w = exp(I(z;ŷ) - I_0), where I(z;ŷ) is estimated from the current batch. Experiments on ImageNet-C and ImageNet-C-bar report that UnInfo outperforms existing CLIP TTA methods (TPT, TDA, ZERO, MTA) and few-shot adaptation baselines, with ablations showing each component contributes and a throughput analysis showing a computational advantage over augmentation-based baselines.

Significance. If the results hold, the paper offers a potentially useful TTA recipe for CLIP under image corruption, with a clear empirical comparison, ablations, sensitivity analysis, efficiency measurements, and released code. The claimed mechanism, however, is that corruptions reduce the information retained in image embeddings and that minimizing a uniformity loss recovers this information. This theoretical framing is not currently supported by the derivation in Section III, and the main experimental comparison is weakened by hyperparameter selection on a subset of the test distribution. The empirical method may still be valuable as a regularized entropy-minimization approach, but the paper's central explanatory claim needs substantial revision.

major comments (3)
  1. [Section III, Eq. (4)] The derivation of the information-theoretic motivation is invalid. In the Markov chain X -> X' -> Z, X' is a stochastic corruption, so the image encoder, while deterministic given X', does not make Z deterministic given X. Consequently H(Z|X) = H(f_img(X')|X) is not zero and is not in general a constant; it can grow with the entropy of the corruption process. The data processing inequality I(X;Z) <= I(X;X') therefore does not imply that H(Z) becomes smaller under corruption. Moreover, the uniformity loss in Eq. (3) is a pairwise RBF statistic of the marginal distribution of embeddings, not an estimator of H(Z) or I(X;Z). A deterministic encoder can map noisy inputs to near-uniform embeddings that carry little class-relevant information, so minimizing this loss does not by itself establish that classification-relevant information is retained. The central claim that UnInfo improves accuracy 'by retaining information in terms of uniformity' is not supported by the presented theory. Please either provide a valid derivation or reframe the uniformity loss as an empirical regularizer that prevents entropy-minimization collapse.
  2. [Section V-B] The hyperparameters λ=1 and I_0=3 are selected using 'a few corruption types in ImageNet-C', which is part of the test distribution. In the TTA protocol, the test distribution is unknown and labels are unavailable; selecting hyperparameters on a subset of the evaluation set and then reporting the average over all corruptions inflates the reported gains and breaks the evaluation protocol. This issue is load-bearing because the information-balancing mechanism depends on the threshold I_0, and the claim that the balancing adaptively detects the current regime is weakened if I_0 is tuned on the test distribution. Please specify exactly which corruption types were used for selection, report results excluding those types, or choose hyperparameters using a disjoint validation set that is not part of the test benchmark.
  3. [Abstract and Table III] The claim that existing TTA methods 'fail to adapt to distribution shifts caused by sensor degradation' is too strong. Table III shows that TPT improves mean accuracy from 23.09 to 25.07 and TDA from 23.09 to 25.54, with both methods improving on most individual corruption types. The results show limited or inconsistent gains, not a complete failure. The text should be revised to reflect this more accurately, since the overstatement affects the paper's motivation and framing.
minor comments (6)
  1. [Section IV-B, Eq. (10)] The notation in Eq. (10) is misleading: the first equality writes I(z;ŷ) = H(z) - H(z|ŷ), but the computed quantities are H(ŷ) - H(ŷ|z) using the discrete entropy of the predicted-label distribution. Since mutual information is symmetric this is numerically correct, but the notation should be fixed to avoid confusion.
  2. [Section V-B] The LoRA rank and alpha are said to be set to default values, but the actual values are not reported. Please specify these hyperparameters for reproducibility.
  3. [Section V-D4, Table VII] The discussion emphasizes UnInfo's throughput advantage, but Table VII shows that UnInfo also has the second-highest GPU memory usage (11736 MiB). This trade-off should be acknowledged explicitly.
  4. [Section III and Section V-A] The preliminary experiment in Section III uses a ViT-B/16 CLIP trained on LAION, while the main experiments use a ViT-B/16 CLIP trained on DataComp-1B. The choice of different checkpoints should be clarified earlier, since the preliminary observations motivate the method but are not directly obtained with the model used in the main evaluation.
  5. [Section V-D5, Figure 4] The qualitative claim that uniformity is improved after TTA is based on a 2D spherical PCA visualization. Reporting the numerical uniformity loss before and after adaptation would strengthen this claim.
  6. [Section V-B] The sensitivity of the EMA momentum parameter m is not reported, even though the ablation shows that the EMA teacher plays a large role in preventing collapse. A sensitivity analysis for m would be useful.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: UnInfo's accuracy gains are measured on external corruption benchmarks, and the theoretical flaw in Eq. (4) is a correctness issue, not a circular reduction.

full rationale

I walked the paper's derivation chain. The central mechanism is the uniformity loss Lunif (Eq. 7) added to entropy minimization and EMA distillation (Eq. 15). The paper does not define uniformity in terms of its own final accuracy; it imports the uniformity loss as an information-related regularizer from external work [45, 53] and tests it empirically via ablations (Tables V and VI). The theoretical argument in Section III, Eq. (4), contains a technical flaw: for a stochastic corruption chain X -> X' -> Z with a deterministic encoder, H(Z|X) is not constant, and the differential entropy H(Z) is not the same quantity as the uniformity loss of Eq. (3). But a false equation is not a circular reduction: no fitted parameter is renamed as a prediction, and the method's improvements are measured on the actual ImageNet-C and ImageNet-C-bar test sets. The hyperparameters λ and I0 are chosen using a few ImageNet-C corruption types and then held fixed; this is a benchmark-tuning weakness, not circularity, because the reported accuracies are measured results rather than quantities forced by the fitted values. Self-citations [23, 42, 50] appear only as related work and are not load-bearing for the central claim. I therefore find no step that reduces, by the paper's own equations or by self-citation, to its own inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The method is built from known components (LoRA, entropy minimization, uniformity loss, EMA distillation) and introduces one adaptive weighting scheme whose threshold I_0 and weight λ are fitted on a subset of the benchmark used for evaluation. No new physical or conceptual entities are introduced. The main axiom burden is the assumed link between the uniformity loss and information retention, plus the reliability of the batch mutual information estimate.

free parameters (6)
  • λ (uniformity loss weight) = 1
    Weight of uniformity loss in Eq. (8). Set via tuning on a few corruption types of ImageNet-C (Section V-B).
  • I_0 (mutual information threshold) = 3
    Threshold in Eq. (9) controlling entropy/uniformity balance. Tuned on a few ImageNet-C corruption types; sensitivity shows best near 2.75.
  • Learning rate and weight decay = 0.001, 0.01
    AdamW hyperparameters chosen alongside λ and I_0 on a subset of ImageNet-C.
  • EMA momentum m = 0.001
    Teacher momentum in Eq. (13); set without reported ablations.
  • LoRA rank and alpha = default
    Taken as defaults from Zanella & Ayed [60]; no tuning reported.
  • Batch size = 64
    Batch size for TTA; affects the uniformity and mutual information estimates.
assumptions (5)
  • standard math The image corruption process forms a Markov chain X → X′ → Z, so the data processing inequality gives I(X;Z) ≤ I(X;X′).
    Section III, Eq. (4). Standard DPI; the interpretation that smaller I(X;X′) implies smaller H(Z) relies on this.
  • domain assumption The uniformity loss (Eq. 3) is a faithful proxy for the amount of input information retained in image embeddings.
    Section III and Section IV-A. The paper equates reduced uniformity with reduced information without a formal equivalence between the Rényi-2 divergence and differential entropy.
  • domain assumption The batch-based mutual information estimate in Eq. (10) reliably identifies the TTA regime.
    Section IV-B. Uses plug-in estimates of marginal and conditional entropy; reliability under severe corruption is not analyzed.
  • domain assumption Updating only the image encoder via LoRA, keeping text embeddings fixed, is sufficient for TTA under corruption.
    Section IV-C. The method assumes text prototypes remain valid and only image embeddings are corrupted.
  • domain assumption The EMA teacher provides a stable, non-collapsing target during TTA.
    Section IV-D. Ablations confirm L_ent alone collapses to 0.10 accuracy, so the teacher is load-bearing for stability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Uniformity First: Uniformity-aware Test-time Adaptation of Vision-language Models against Image Corruption." pith.science (2026). https://pith.science/paper/S2FED67D

@misc{pith2026250512912,
  author       = {Pith},
  title        = {Pith review of: Uniformity First: Uniformity-aware Test-time Adaptation of Vision-language Models against Image Corruption},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S2FED67D}},
  note         = {Machine review of arXiv:2505.12912}
}
read the original abstract

Pre-trained vision-language models such as contrastive language-image pre-training (CLIP) have demonstrated a remarkable generalizability, which has enabled a wide range of applications represented by zero-shot classification. However, vision-language models still suffer when they face datasets with large gaps from training ones, i.e., distribution shifts. We found that CLIP is especially vulnerable to sensor degradation, a type of realistic distribution shift caused by sensor conditions such as weather, light, or noise. Collecting a new dataset from a test distribution for fine-tuning highly costs since sensor degradation occurs unexpectedly and has a range of variety. Thus, we investigate test-time adaptation (TTA) of zero-shot classification, which enables on-the-fly adaptation to the test distribution with unlabeled test data. Existing TTA methods for CLIP mainly focus on modifying image and text embeddings or predictions to address distribution shifts. Although these methods can adapt to domain shifts, such as fine-grained labels spaces or different renditions in input images, they fail to adapt to distribution shifts caused by sensor degradation. We found that this is because image embeddings are "corrupted" in terms of uniformity, a measure related to the amount of information. To make models robust to sensor degradation, we propose a novel method called uniformity-aware information-balanced TTA (UnInfo). To address the corruption of image embeddings, we introduce uniformity-aware confidence maximization, information-aware loss balancing, and knowledge distillation from the exponential moving average (EMA) teacher. Through experiments, we demonstrate that our UnInfo improves accuracy under sensor degradation by retaining information in terms of uniformity.

Figures

Figures reproduced from arXiv: 2505.12912 by the authors.

Figure 1
Figure 1. Overview of Uniformity-aware Information-balanced Test-time Adaptation (UnInfo). [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. Evolution of the information-aware loss balancing weights. The weights are adaptively assigned to the entropy and uniformity losses by the difficulty [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Spherical PCA [69] visualization of image ( [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Adapting Vision-Language Models Without Labels: A Comprehensive Survey

    cs.LG 2025-08 conditional novelty 5.0 of 10

    A survey that organizes unsupervised vision-language model adaptation by unlabeled-data availability into four paradigms: data-free transfer, domain transfer, episodic test-time, and online test-time adaptation.

Reference graph

Works this paper leans on

71 extracted references · 52 canonical work pages · cited by 1 Pith paper

  1. [1]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clarket al., “Learning transferable visual models from natural language supervision,” inInternational conference on machine learning. PMLR, 2021, pp. 8748–8763

    2021

  2. [2]

    Scaling up visual and vision-language representation learning with noisy text supervision,

    C. Jia, Y . Yang, Y . Xia, Y .-T. Chen, Z. Parekh, H. Pham, Q. Le, Y .-H. Sung, Z. Li, and T. Duerig, “Scaling up visual and vision-language representation learning with noisy text supervision,” inInternational conference on machine learning. PMLR, 2021, pp. 4904–4916

    2021

  3. [3]

    Improving zero-shot generalization and robustness of multi-modal models,

    Y . Ge, J. Ren, A. Gallagher, Y . Wang, M.-H. Yang, H. Adam, L. Itti, B. Lakshminarayanan, and J. Zhao, “Improving zero-shot generalization and robustness of multi-modal models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 11 093–11 101

    work page 2023

  4. [4]

    Improving zero- shot generalization for clip with synthesized prompts,

    Z. Wang, J. Liang, R. He, N. Xu, Z. Wang, and T. Tan, “Improving zero- shot generalization for clip with synthesized prompts,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 3032–3042

    work page 2023

  5. [5]

    Effective con- ditioned and composed image retrieval combining clip-based features,

    A. Baldrati, M. Bertini, T. Uricchio, and A. Del Bimbo, “Effective con- ditioned and composed image retrieval combining clip-based features,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 21 466–21 474. 10

    work page 2022

  6. [6]

    Clip2video: Mastering video- text retrieval via image clip,

    H. Fang, P. Xiong, L. Xu, and Y . Chen, “Clip2video: Mastering video- text retrieval via image clip,”arXiv preprint arXiv:2106.11097, 2021

    arXiv 2021

  7. [7]

    Styleclip: Text-driven manipulation of stylegan imagery,

    O. Patashnik, Z. Wu, E. Shechtman, D. Cohen-Or, and D. Lischinski, “Styleclip: Text-driven manipulation of stylegan imagery,” inProceed- ings of the IEEE/CVF international conference on computer vision, 2021, pp. 2085–2094

    work page 2021

  8. [8]

    Hierarchical text-conditional image generation with clip latents,

    A. Ramesh, P. Dhariwal, A. Nichol, C. Chu, and M. Chen, “Hierarchical text-conditional image generation with clip latents,”arXiv preprint arXiv:2204.06125, vol. 1, no. 2, p. 3, 2022

    arXiv 2022

Show all 71 references
  1. [9]

    OpenCLIP,

    G. Ilharco, M. Wortsman, R. Wightman, C. Gordon, N. Carlini, R. Taori, A. Dave, V . Shankar, H. Namkoong, J. Miller, H. Hajishirzi, A. Farhadi, and L. Schmidt, “OpenCLIP,” Jul. 2021

    2021

  2. [10]

    Reproducible scaling laws for contrastive language-image learning,

    M. Cherti, R. Beaumont, R. Wightman, M. Wortsman, G. Ilharco, C. Gordon, C. Schuhmann, L. Schmidt, and J. Jitsev, “Reproducible scaling laws for contrastive language-image learning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp....

    2023

  3. [11]

    LAION-5b: An open large-scale dataset for training next generation image-text models,

    C. Schuhmann, R. Beaumont, R. Vencu, C. W. Gordon, R. Wight- man, M. Cherti, T. Coombes, A. Katta, C. Mullis, M. Wortsman, P. Schramowski, S. R. Kundurthy, K. Crowson, L. Schmidt, R. Kacz- marczyk, and J. Jitsev, “LAION-5b: An open large-scale dataset for training next generat...

    2022

  4. [12]

    Tip-Adapter: Training-free Adaption of CLIP for Few-shot Classification,

    R. Zhang, R. Fang, W. Zhang, P. Gao, K. Li, J. Dai, Y . Qiao, and H. Li, “Tip-Adapter: Training-free Adaption of CLIP for Few-shot Classification,” inEuropean Conference on Computer Vision (ECCV), 2022

    2022

  5. [13]

    Lp++: A surprisingly strong linear probe for few-shot clip,

    Y . Huang, F. Shakeri, J. Dolz, M. Boudiaf, H. Bahig, and I. Ben Ayed, “Lp++: A surprisingly strong linear probe for few-shot clip,” inPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 773–23 782

    2024

  6. [14]

    PLOT: Prompt learning with optimal transport for vision-language models,

    G. Chen, W. Yao, X. Song, X. Li, Y . Rao, and K. Zhang, “PLOT: Prompt learning with optimal transport for vision-language models,” inThe Eleventh International Conference on Learning Representations, 2023

    2023

  7. [15]

    Test-time prompt tuning for zero-shot generalization in vision-language models,

    S. Manli, N. Weili, H. De-An, Y . Zhiding, G. Tom, A. Anima, and X. Chaowei, “Test-time prompt tuning for zero-shot generalization in vision-language models,” inNeurIPS, 2022

    2022

  8. [16]

    Test-time distri- bution normalization for contrastively learned visual-language models,

    Y . Zhou, J. Ren, F. Li, R. Zabih, and S. N. Lim, “Test-time distri- bution normalization for contrastively learned visual-language models,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  9. [17]

    Efficient Test-Time Adaptation of Vision-Language Models,

    A. Karmanov, D. Guan, S. Lu, A. El Saddik, and E. Xing, “Efficient Test-Time Adaptation of Vision-Language Models,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 14 162–14 171

    2024

  10. [18]

    Dual memory networks: A versatile adaptation approach for vision-language models,

    Y . Zhang, W. Zhu, H. Tang, Z. Ma, K. Zhou, and L. Zhang, “Dual memory networks: A versatile adaptation approach for vision-language models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 28 718–28 728

    2024

  11. [19]

    On the test-time zero-shot generalization of vision-language models: Do we really need prompt learning?

    M. Zanella and I. Ben Ayed, “On the test-time zero-shot generalization of vision-language models: Do we really need prompt learning?” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 783–23 793

    2024

  12. [20]

    A hard-to- beat baseline for training-free CLIP-based adaptation,

    Z. Wang, J. Liang, L. Sheng, R. He, Z. Wang, and T. Tan, “A hard-to- beat baseline for training-free CLIP-based adaptation,” inThe Twelfth International Conference on Learning Representations, 2024

    2024

  13. [21]

    Online zero-shot classification with clip,

    Q. Qian and J. Hu, “Online zero-shot classification with clip,” in European Conference on Computer Vision. Springer, 2024, pp. 462– 477

    2024

  14. [22]

    Tent: Fully Test-Time Adaptation by Entropy Minimization,

    D. Wang, E. Shelhamer, S. Liu, B. Olshausen, and T. Darrell, “Tent: Fully Test-Time Adaptation by Entropy Minimization,” inInternational Conference on Learning Representations (ICLR), 2021

    2021

  15. [23]

    Covariance-Aware Feature Alignment with Pre-Computed Source Statistics for Test-Time Adap- tation to Multiple Image Corruptions,

    K. Adachi, S. Yamaguchi, and A. Kumagai, “Covariance-Aware Feature Alignment with Pre-Computed Source Statistics for Test-Time Adap- tation to Multiple Image Corruptions,” in2023 IEEE International Conference on Image Processing (ICIP), 2023, pp. 800–804

    2023

  16. [24]

    A comprehensive survey on test-time adaptation under distribution shifts,

    J. Liang, R. He, and T. Tan, “A comprehensive survey on test-time adaptation under distribution shifts,”International Journal of Computer Vision, pp. 1–34, 2024

    2024

  17. [25]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,”International Journal of Computer Vision (IJCV), 2022

    2022

  18. [26]

    Conditional prompt learning for vision-language models,

    ——, “Conditional prompt learning for vision-language models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 16 816–16 825

    2022

  19. [27]

    Mind the gap: Understanding the modality gap in multi-modal contrastive representation learning,

    V . W. Liang, Y . Zhang, Y . Kwon, S. Yeung, and J. Y . Zou, “Mind the gap: Understanding the modality gap in multi-modal contrastive representation learning,”Advances in Neural Information Processing Systems, vol. 35, pp. 17 612–17 625, 2022

    2022

  20. [28]

    Maple: Multi-modal prompt learning,

    M. U. Khattak, H. Rasheed, M. Maaz, S. Khan, and F. S. Khan, “Maple: Multi-modal prompt learning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 19 113–19 122

    2023

  21. [29]

    Intra-modal proxy learning for zero- shot visual categorization with clip,

    Q. Qian, Y . Xu, and J. Hu, “Intra-modal proxy learning for zero- shot visual categorization with clip,”Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  22. [30]

    Post- pre-training for Modality Alignment in Vision-Language Foundation Models,

    S. Yamaguchi, D. Feng, S. Kanai, K. Adachi, and D. Chijiwa, “Post- pre-training for Modality Alignment in Vision-Language Foundation Models,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2025

    2025

  23. [31]

    The Many Faces of Robustness: A Critical Analysis of Out- of-Distribution Generalization,

    D. Hendrycks, S. Basart, N. Mu, S. Kadavath, F. Wang, E. Dorundo, R. Desai, T. Zhu, S. Parajuli, M. Guo, D. Song, J. Steinhardt, and J. Gilmer, “The Many Faces of Robustness: A Critical Analysis of Out- of-Distribution Generalization,”arXiv preprint arXiv:2006.16241, 2020

    2006 arXiv

  24. [32]

    Do imagenet classifiers generalize to imagenet?

    B. Recht, R. Roelofs, L. Schmidt, and V . Shankar, “Do imagenet classifiers generalize to imagenet?” inInternational Conference on Machine Learning. PMLR, 2019, pp. 5389–5400

    2019

  25. [33]

    Natural Adversarial Examples,

    D. Hendrycks, K. Zhao, S. Basart, J. Steinhardt, and D. Song, “Natural Adversarial Examples,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2021, pp. 15 262–15 271

    2021

  26. [34]

    Learning Robust Global Representations by Penalizing Local Predictive Power,

    H. Wang, S. Ge, Z. Lipton, and E. P. Xing, “Learning Robust Global Representations by Penalizing Local Predictive Power,” inAdvances in Neural Information Processing Systems 32, H. Wallach, H. Larochelle, A. Beygelzimer, F. d'Alch ´e-Buc, E. Fox, and R. Garnett, Eds. Curran As...

    2019

  27. [35]

    Automated flower classification over a large number of classes,

    M.-E. Nilsback and A. Zisserman, “Automated flower classification over a large number of classes,” inIndian Conference on Computer Vision, Graphics and Image Processing, Dec 2008

    2008

  28. [36]

    Cats and dogs,

    O. M. Parkhi, A. Vedaldi, A. Zisserman, and C. V . Jawahar, “Cats and dogs,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2012

    2012

  29. [37]

    Benchmarking Neural Network Ro- bustness to Common Corruptions and Perturbations,

    D. Hendrycks and T. Dietterich, “Benchmarking Neural Network Ro- bustness to Common Corruptions and Perturbations,” inProceedings of the International Conference on Learning Representations (ICLR), 2019

    2019

  30. [38]

    Ar-tta: A simple method for real-world continual test-time adaptation,

    D. S ´ojka, S. Cygert, B. Twardowski, and T. Trzci ´nski, “Ar-tta: A simple method for real-world continual test-time adaptation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 3491–3495

    2023

  31. [39]

    Dark model adaptation: Semantic image segmentation from daytime to nighttime,

    D. Dai and L. V . Gool, “Dark model adaptation: Semantic image segmentation from daytime to nighttime,” in2018 21st International Conference on Intelligent Transportation Systems (ITSC), 2018, pp. 3819–3824

    2018

  32. [40]

    Towards robust cnn-based object detection through augmentation with synthetic rain variations,

    G. V olk, S. M ¨uller, A. v. Bernuth, D. Hospach, and O. Bringmann, “Towards robust cnn-based object detection through augmentation with synthetic rain variations,” in2019 IEEE Intelligent Transportation Systems Conference (ITSC), 2019, pp. 285–292

    2019

  33. [41]

    Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restoration,

    C. Eastwood, I. Mason, C. Williams, and B. Sch ¨olkopf, “Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restoration,” inInternational Conference on Learning Representations, 2022

    2022

  34. [42]

    Test-time similar- ity modification for person re-identification toward temporal distribution shift,

    K. Adachi, S. Enomoto, T. Sasaki, and S. Yamaguchi, “Test-time similar- ity modification for person re-identification toward temporal distribution shift,” in2024 International Joint Conference on Neural Networks (IJCNN), 2024, pp. 1–8

    2024

  35. [43]

    Understanding and improving robustness of vision transform- ers through patch-based negative augmentation,

    Y . Qin, C. Zhang, T. Chen, B. Lakshminarayanan, A. Beutel, and X. Wang, “Understanding and improving robustness of vision transform- ers through patch-based negative augmentation,”Advances in Neural Information Processing Systems, vol. 35, pp. 16 276–16 289, 2022

    2022

  36. [44]

    On interaction between augmenta- tions and corruptions in natural corruption robustness,

    E. Mintun, A. Kirillov, and S. Xie, “On interaction between augmenta- tions and corruptions in natural corruption robustness,” inAdvances in Neural Information Processing Systems, A. Beygelzimer, Y . Dauphin, P. Liang, and J. W. Vaughan, Eds., 2021

    2021

  37. [45]

    Understanding contrastive representation learning through alignment and uniformity on the hypersphere,

    T. Wang and P. Isola, “Understanding contrastive representation learning through alignment and uniformity on the hypersphere,” inInternational conference on machine learning. PMLR, 2020, pp. 9929–9939

    2020

  38. [46]

    LoRA: Low-rank adaptation of large language models,

    E. J. Hu, yelong shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” inInternational Conference on Learning Representations, 2022

    2022

  39. [47]

    Bayesian Adaptation for Covariate Shift,

    A. Zhou and S. Levine, “Bayesian Adaptation for Covariate Shift,” Advances in Neural Information Processing Systems, vol. 34, 2021. 11

    2021

  40. [48]

    Ef- ficient test-time model adaptation without forgetting,

    S. Niu, J. Wu, Y . Zhang, Y . Chen, S. Zheng, P. Zhao, and M. Tan, “Ef- ficient test-time model adaptation without forgetting,” inInternational conference on machine learning. PMLR, 2022, pp. 16 888–16 905

    2022

  41. [49]

    Memo: Test time robustness via adaptation and augmentation,

    M. Zhang, S. Levine, and C. Finn, “Memo: Test time robustness via adaptation and augmentation,”Advances in neural information process- ing systems, vol. 35, pp. 38 629–38 642, 2022

    2022

  42. [50]

    Test-time adaptation meets image enhancement: Improving accuracy via uncertainty-aware logit switching,

    S. Enomoto, N. Hasegawa, K. Adachi, T. Sasaki, S. Yamaguchi, S. Suzuki, and T. Eda, “Test-time adaptation meets image enhancement: Improving accuracy via uncertainty-aware logit switching,” in2024 International Joint Conference on Neural Networks (IJCNN), 2024, pp. 1–8

    2024

  43. [51]

    Gpt-4o system card,

    A. Hurst, A. Lerer, A. P. Goucher, A. Perelman, A. Ramesh, A. Clark, A. Ostrow, A. Welihinda, A. Hayes, A. Radfordet al., “Gpt-4o system card,”arXiv preprint arXiv:2410.21276, 2024

    2024 arXiv

  44. [52]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” inThe IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2009, pp. 248–255

    2009

  45. [53]

    Representation learning with contrastive predictive coding,

    A. v. d. Oord, Y . Li, and O. Vinyals, “Representation learning with contrastive predictive coding,”arXiv preprint arXiv:1807.03748, 2018

    2018 arXiv

  46. [54]

    Unsupervised Classifiers, Mutual Information and 'Phantom Targets,

    J. Bridle, A. Heading, and D. MacKay, “Unsupervised Classifiers, Mutual Information and 'Phantom Targets,” inAdvances in Neural Infor- mation Processing Systems, J. Moody, S. Hanson, and R. P. Lippmann, Eds., vol. 4. Morgan-Kaufmann, 1992

    1992

  47. [55]

    Discriminative Clustering by Regularized Information Maximization,

    A. Krause, P. Perona, and R. Gomes, “Discriminative Clustering by Regularized Information Maximization,” inAdvances in Neural Information Processing Systems, J. Lafferty, C. Williams, J. Shawe- Taylor, R. Zemel, and A. Culotta, Eds., vol. 23. Curran Associates, Inc., 2010. [On...

    2010

  48. [56]

    Information-theoretical learning of discriminative clusters for unsupervised domain adaptation,

    Y . Shi and F. Sha, “Information-theoretical learning of discriminative clusters for unsupervised domain adaptation,” inInternational Confer- ence on Machine Learning (ICML), 2012

    2012

  49. [57]

    Learning discrete representations via information maximizing self-augmented training,

    W. Hu, T. Miyato, S. Tokui, E. Matsumoto, and M. Sugiyama, “Learning discrete representations via information maximizing self-augmented training,” inInternational Conference on Machine Learning (ICML), ser. Proceedings of Machine Learning Research, D. Precup and Y . W. Teh, Ed...

    2017

  50. [58]

    Padclip: Pseudo-labeling with adaptive debiasing in clip for unsupervised domain adaptation,

    Z. Lai, N. Vesdapunt, N. Zhou, J. Wu, C. P. Huynh, X. Li, K. K. Fu, and C.-N. Chuah, “Padclip: Pseudo-labeling with adaptive debiasing in clip for unsupervised domain adaptation,” inProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), October 2023, p...

    2023

  51. [59]

    HVCLIP: High-dimensional vector in CLIP for unsupervised domain adaptation,

    N. Vesdapunt, K. K. Fu, Y . Wu, X. Zhang, and P. Natarajan, “HVCLIP: High-dimensional vector in CLIP for unsupervised domain adaptation,” inEuropean Conference on Computer Vision (ECCV), 2024, pp. 36–54

    2024

  52. [60]

    Low-rank few-shot adaptation of vision- language models,

    M. Zanella and I. Ben Ayed, “Low-rank few-shot adaptation of vision- language models,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, June 2024, pp. 1593–1603

    2024

  53. [61]

    Continual test-time domain adaptation,

    Q. Wang, O. Fink, L. Van Gool, and D. Dai, “Continual test-time domain adaptation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 7201–7211

    2022

  54. [62]

    Visual prompt tuning for test-time domain adaptation,

    Y . Gao, X. Shi, Y . Zhu, H. Wang, Z. Tang, X. Zhou, M. Li, and D. N. Metaxas, “Visual prompt tuning for test-time domain adaptation,”arXiv preprint arXiv:2210.04831, 2022

    2022 arXiv

  55. [63]

    Robust mean teacher for continual and gradual test-time adaptation,

    M. D ¨obler, R. A. Marsden, and B. Yang, “Robust mean teacher for continual and gradual test-time adaptation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 7704–7714

    2023

  56. [64]

    Continual test-time domain adaptation via dynamic sample selection,

    Y . Wang, J. Hong, A. Cheraghian, S. Rahman, D. Ahmedt-Aristizabal, L. Petersson, and M. Harandi, “Continual test-time domain adaptation via dynamic sample selection,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), January 2024, pp. ...

    2024

  57. [65]

    Datacomp: In search of the next generation of multimodal datasets,

    S. Y . Gadre, G. Ilharco, A. Fang, J. Hayase, G. Smyrnis, T. Nguyen, R. Marten, M. Wortsman, D. Ghosh, J. Zhanget al., “Datacomp: In search of the next generation of multimodal datasets,”Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  58. [66]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” inInternational Conference on Learning Representations, 2019

    2019

  59. [67]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,

    K. He, X. Zhang, S. Ren, and J. Sun, “Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,” in Proceedings of the IEEE International Conference on Computer Vision (ICCV), December 2015

    2015

  60. [68]

    Frustratingly easy test-time adaptation of vision-language models,

    M. Farina, G. Franchi, G. Iacca, M. Mancini, and E. Ricci, “Frustratingly easy test-time adaptation of vision-language models,” inThe Thirty- eighth Annual Conference on Neural Information Processing Systems, 2024

    2024

  61. [69]

    Spherical principal component analysis,

    K. Liu, Q. Li, H. Wang, and G. Tang, “Spherical principal component analysis,” inProceedings of the 2019 SIAM International Conference on Data Mining. SIAM, 2019, pp. 387–395

    2019

  62. [70]

    Microsoft coco captions: Data collection and evaluation server,

    X. Chen, H. Fang, T.-Y . Lin, R. Vedantam, S. Gupta, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco captions: Data collection and evaluation server,”arXiv preprint arXiv:1504.00325, 2015

    2015 arXiv

  63. [71]

    C-TPT: Calibrated Test-Time Prompt Tuning for Vision-Language Models via Text Feature Dispersion,

    H. S. Yoon, E. Yoon, J. T. J. Tee, M. Hasegawa-Johnson, Y . Li, and C. D. Yoo, “C-TPT: Calibrated Test-Time Prompt Tuning for Vision-Language Models via Text Feature Dispersion,” inInternational Conference on Learning Representations (ICLR), 2024. APPENDIX TEXTPROMPTENSEMBLE H...

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.