Pith. sign in

REVIEW 3 major objections 2 minor 74 references

Clustering-based aggregate value regression

T0 review · 3 major / 2 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Aggregate Value Regression combines many linear models into one and claims that the cluster count — not error from unsupervised clustering — sets the forecast's bias-variance trade-off even under misspecification.

desk verdict The abstract promises a useful bias-variance theory for clustered aggregate regression, but the supplied full text is an unrelated vision paper, so there is nothing to referee in this submission. read the letter →

arxiv 2508.15567 v1 pith:X562XOFC submitted 2025-08-21 stat.ME

Kei Hirose , Hidetoshi Matsui , Hiroki Masuda This is my paper
classification stat.ME MSC 62J0562H30
keywords aggregateforecastinglinearregressionhierarchicalclusteringbias-variancetrade-offmodelmisspecificationcomplexityelectricitydemandoverparameterization
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

This paper aims to make forecasting of aggregate quantities — a regional total, say, rather than each meter or customer — a first-class statistical problem. It proposes Aggregate Value Regression (AVR), which combines all per-unit linear regression models into a single estimation problem, and AVR-C, a hierarchical-clustering version that groups the regressions to avoid estimating an unwieldy number of parameters. The central claim is that under a misspecified model, the number of clusters plays the role of model complexity: fewer clusters shrink variance at the cost of bias, more clusters do the opposite, so cluster choice becomes a statistical trade-off rather than an unsupervised clustering output. If the claim holds, practitioners forecasting totals (electricity demand is the worked example) gain a principled way to select the aggregation level, with the trade-off demonstrated by Monte Carlo experiments and an electricity demand analysis.

What carries the argument

The central object is the cluster of regression models itself. AVR-C builds a hierarchical clustering of the per-unit linear regression models and fits the aggregate-value regression inside each cluster; the number of clusters then indexes model complexity. The mechanism doing the work is the bias-variance decomposition of the aggregate forecast error under a misspecified model: each added cluster buys lower bias (more flexibility to fit unit heterogeneity) at the price of higher variance (more parameters estimated from the same data), so the error curve across cluster counts is trade-off-shaped.

What would settle it

Simulate units whose true coefficients come from a small number of latent groups (for example, three mixture components), then fit AVR-C at every cluster count from one up to many. The theory predicts aggregate test error is minimized near the count that balances bias and variance, not at the true number of groups. If the error curve is monotone decreasing with cluster count, or if its minimum tracks the true number of groups instead of the bias-variance balance, the claimed complexity interpretation fails.

Watch

Extended reading notes

Core claim

On the paper's terms, the discovery is that aggregate forecast error can be organized by a single integer: the number of clusters into which unit-level regression models are grouped. AVR estimates one regression system for all units; with many units this system is overparameterized, so AVR-C imposes hierarchical clustering and re-estimates within each cluster. The stated contribution is a bias-variance trade-off theory under model misspecification: the number of clusters characterizes model complexity, turning clustering into a complexity-control knob rather than an unsupervised description tool. Monte Carlo experiments and an electricity-demand analysis demonstrate the trade-off.

Load-bearing premise

The trade-off claim depends on the model being misspecified in a way that still yields a well-defined bias-variance decomposition of the aggregate error, and on hierarchical clustering of the regressions preserving that decomposition — if either fails, cluster count does not actually govern the forecast error.

Editorial extensions

If this is right

  • Cluster count becomes a tunable model-selection parameter: the forecaster picks the aggregation level where estimated test error is minimized rather than accepting clusters as a fixed data-driven output.
  • Direct aggregate estimation tailors the fitted model to the total, so unit-level parameters are learned with the aggregate loss in mind — a regime where fitting each unit separately can be systematically off-target.
  • The method supplies a statistical rationale for how coarse to make forecasts (regional versus per-meter), guided by the misspecification-aware trade-off.
  • Electricity demand forecasting, and analogous aggregate problems such as network traffic or regional sales, gain a concrete procedure for jointly estimating many regression models without drowning in parameters.
  • Because the theory is stated under misspecification, the result addresses realistic settings where the linear model is known to be approximate, broadening where the trade-off logic applies.

Reading between the lines

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

  • An immediate testable extension: on any aggregate data set with natural unit structure, plot aggregate test error against cluster count; the theory predicts a U-shape whose minimum identifies the operational complexity level — a diagnostic the paper's Monte Carlo supports but does not codify into a formal rule.
  • The same complexity-as-cluster-count logic may transfer beyond linear regression — for instance to generalized linear or quantile unit-level models — where the bias-variance split would need re-derivation but the clustering idea is agnostic to the unit model.
  • A consequence the paper does not spell out: under misspecification, the optimal number of clusters need not coincide with any true underlying grouping of units; it is a purely forecast-optimal choice, so interpreting the clusters substantively could mislead.
  • The trade-off framing suggests a practical rule of thumb: when unit heterogeneity is suspected but unmodeled, start with many clusters and coarsen only until variance savings outweigh bias cost — a heuristic that could be validated against the paper's Monte Carlo design on real load data.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 2 minor

Summary. The abstract of this submission introduces Aggregate Value Regression (AVR), a method for forecasting aggregate quantities by combining unit-level linear regression models, and AVR-C, a hierarchical-clustering variant intended to control overparameterization. The abstract claims a novel bias-variance trade-off under model misspecification, with the number of clusters serving as a model-complexity parameter, and states that Monte Carlo simulation and an electricity-demand application demonstrate the theory. The supplied full text, however, is an unrelated computer-vision paper on test-time adaptation (ADAPT, arXiv:2508.15568v8). None of the AVR/AVR-C definitions, derivations, simulation protocol, or empirical analysis appears in the manuscript body. The central claim is therefore unverifiable from the submitted document.

Significance. If the claimed bias-variance trade-off for clustered aggregate-value regression is correct, it would be a useful contribution to statistical forecasting: it would connect the number of clusters to forecast error in a misspecified-model setting and offer a principled complexity choice for AVR-C. The manuscript as submitted provides no way to assess this contribution. There are no equations, no derivations, no simulation details, no numerical results, and no machine-checked or reproducible artifacts. The abstract's claims are not falsifiable because the method is not defined at the level of mathematical specification. The mismatch between the abstract and the body is a blocking problem, not a matter of presentation.

major comments (3)
  1. [Full text (manuscript body)] The body of the submitted manuscript is a NeurIPS-style vision paper on ADAPT (test-time adaptation via probabilistic Gaussian alignment); it contains no mention of AVR, AVR-C, aggregate forecasting, linear regression models, clustering of regressions, or electricity demand. None of the abstract's advertised technical content — the definition of AVR, the combination rule, the cluster construction, the bias-variance decomposition, the Monte Carlo design, or the empirical analysis — is present. This is load-bearing because the central claim of the paper (a new bias-variance trade-off theory) cannot be checked in any way from this submission.
  2. [Abstract] Even treating the abstract as the only content, AVR-C is not specified enough to support the claimed theory. 'Combining all regression models into a single model' does not state the form of the combined estimator, the target of aggregation, or the loss function. 'Hierarchical clustering technique' does not specify the dissimilarity measure between regression models, the linkage criterion, or how cluster-level AVR forecasts are combined into an aggregate forecast. Without these ingredients, the statement that the number of clusters characterizes model complexity is a qualitative claim, not a derivable result.
  3. [Abstract, bias-variance trade-off claim] The trade-off is asserted 'under the assumption of a misspecified model,' but the nature and degree of misspecification are never defined. It is unclear whether bias is taken with respect to the true conditional expectation or the best linear approximation, what covariance structure is assumed across units, or over which sampling distribution the forecast error is evaluated. A bias-variance theory with these ingredients unspecified is not assessable. This is a load-bearing gap, not a missing detail.
minor comments (2)
  1. [Abstract] The two claimed demonstrations — Monte Carlo simulation and electricity-demand forecasting — are mentioned without any numerical findings, effect sizes, or error bars, so the reader cannot gauge the strength of the evidence.
  2. [Abstract] The phrase 'to our knowledge, statistical learning specifically for forecasting aggregate values has not yet been well-established' would benefit from engagement with the existing forecast-reconciliation and multi-task learning literature in a resubmitted manuscript.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity detectable: the supplied full text is a different paper (ADAPT), so the AVR-C derivation chain is absent; absence of support is not circularity.

full rationale

The manuscript as supplied consists of an abstract for 'Clustering-based aggregate value regression' (AVR-C) and a full text that is actually a different paper: 'Backpropagation-Free Test-Time Adaptation via Probabilistic Gaussian Alignment' (arXiv:2508.15568v8), a NeurIPS 2025 submission on test-time adaptation. There is therefore no derivation chain for AVR-C present in the submitted material. The abstract's central claim—that AVR-C introduces a novel bias-variance trade-off theory under model misspecification, with cluster count characterizing model complexity—is stated without any accompanying equations, definitions, theorem statements, simulation protocol, or empirical analysis. I cannot exhibit a specific reduction of any prediction to its inputs because no equations or fitted parameters for AVR-C appear in the body. The Monte Carlo simulation and electricity demand forecasting are mentioned in the abstract only as validation activities, not as components of a derivation that could be circular. No self-citations, imported uniqueness theorems, or ansatz-via-citation patterns are present in the supplied text. Under the hard rule that circularity may be claimed only when the paper itself shows the reduction, the correct finding is no circularity. This should not be mistaken for verification of the AVR-C claims: the mismatch between abstract and body is a serious document-integrity and verifiability problem, but it is not a circularity problem.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

Provisional ledger reconstructed from the abstract only. The supplied full text is a different paper, so it was not used as evidence. Two candidate free parameters are visible at abstract level: the number of clusters (explicitly designated as the complexity axis) and the unspecified rule for combining regression models in AVR. Three axioms are identified: explicit model misspecification, the extension of bias-variance decomposition to aggregate forecasts, and the fidelity of the cluster reduction. No invented entities (new particles, forces, dimensions, or quantities) are proposed. This ledger must be re-audited against the actual full text.

free parameters (2)
  • Number of clusters in AVR-C = not reported in abstract
    The abstract designates cluster count as the model-complexity axis ('the number of clusters characterizes model complexity'). Its selection is a tuning decision; the bias-variance theory is expressed in terms of it, and the abstract does not state how it is chosen.
  • Combination rule or weights for merging regression models in AVR = unspecified in abstract
    AVR is 'constructed by combining all regression models into a single model' (abstract). Whether the combination is equal-weight averaging (no free parameter) or a weighted/estimated scheme (free parameters) cannot be determined from the abstract; if estimated from data, the weights are fitted values the central claim depends on.
assumptions (3)
  • domain assumption The true data-generating process lies outside the fitted linear model class (model misspecification).
    Stated in the abstract as the setting for the AVR-C theory: 'under the assumption of a misspecified model.' The theory's validity depends on the nature and degree of misspecification, which the abstract does not specify.
  • standard math The classical bias-variance decomposition extends to aggregated (summed) predictions of combined regression models.
    The claimed 'novel bias-variance trade-off theory' generalizes a classical decomposition to the aggregate forecast and to misspecified models; this extension is asserted in the abstract rather than demonstrated there.
  • domain assumption Hierarchical clustering of regression models is a faithful complexity reduction, preserving the aggregate forecast's bias-variance structure.
    AVR-C 'constructs several clusters of regression models' to escape overparameterization (abstract). That cluster-level aggregation approximates full AVR well enough for the trade-off theory to govern it is an implicit modeling premise, unverifiable from the abstract.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Clustering-based aggregate value regression." pith.science (2026). https://pith.science/paper/X562XOFC

@misc{pith2026250815567,
  author       = {Pith},
  title        = {Pith review of: Clustering-based aggregate value regression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X562XOFC}},
  note         = {Machine review of arXiv:2508.15567}
}
read the original abstract

In various practical situations, forecasting of aggregate values rather than individual ones is often our main focus. For instance, electricity companies are interested in forecasting the total electricity demand in a specific region to ensure reliable grid operation and resource allocation. However, to our knowledge, statistical learning specifically for forecasting aggregate values has not yet been well-established. In particular, the relationship between forecast error and the number of clusters has not been well studied, as clustering is usually treated as unsupervised learning. This study introduces a novel forecasting method specifically focused on the aggregate values in the linear regression model. We call it the Aggregate Value Regression (AVR), and it is constructed by combining all regression models into a single model. With the AVR, we must estimate a huge number of parameters when the number of regression models to be combined is large, resulting in overparameterization. To address the overparameterization issue, we introduce a hierarchical clustering technique, referred to as AVR-C (C stands for clustering). In this approach, several clusters of regression models are constructed, and the AVR is performed within each cluster. The AVR-C introduces a novel bias-variance trade-off theory under the assumption of a misspecified model. In this framework, the number of clusters characterizes model complexity. Monte Carlo simulation is conducted to investigate the behavior of training and test errors of our proposed clustering technique. The bias-variance trade-off theory is also demonstrated through the analysis of electricity demand forecasting.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

74 extracted references · 5 canonical work pages

  1. [1]

    Food-101–Mining Discriminative Components with Random Forests

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101–Mining Discriminative Components with Random Forests. InECCV, 2014

    2014

  2. [2]

    Information maximization for few-shot learning

    Malik Boudiaf, Imtiaz Ziko, Jérôme Rony, José Dolz, Pablo Piantanida, and Ismail Ben Ayed. Information maximization for few-shot learning. InNeurIPS, 2020

    2020

  3. [3]

    Describing Textures in the Wild

    Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and Andrea Vedaldi. Describing Textures in the Wild. InCVPR, 2014

    2014

  4. [4]

    Imagenet: A Large-Scale Hierarchical Image Database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A Large-Scale Hierarchical Image Database. InCVPR, 2009

    2009

  5. [5]

    A normality test for multivariate dependent samples.Signal Processing, 201:108705, 2022

    Sara El Bouch, Olivier Michel, and Pierre Comon. A normality test for multivariate dependent samples.Signal Processing, 201:108705, 2022

    2022

  6. [6]

    Joint normality test via two-dimensional projection

    Sara ElBouch, Olivier JJ Michel, and Pierre Comon. Joint normality test via two-dimensional projection. InICASSP, 2022

    2022

  7. [7]

    Frus- tratingly easy test-time adaptation of vision-language models

    Matteo Farina, Gianni Franchi, Giovanni Iacca, Massimiliano Mancini, and Elisa Ricci. Frus- tratingly easy test-time adaptation of vision-language models. InNeurIPS, 2024

    2024

  8. [8]

    Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories

    Li Fei-Fei, Rob Fergus, and Pietro Perona. Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories. In CVPRW, 2004

    2004

Show all 74 references
  1. [9]

    Diverse data augmenta- tion with diffusions for effective test-time prompt tuning

    Chun-Mei Feng, Kai Yu, Yong Liu, Salman Khan, and Wangmeng Zuo. Diverse data augmenta- tion with diffusions for effective test-time prompt tuning. InICCV, 2023

    2023

  2. [10]

    Online gaussian test-time adaptation of vision-language models.arXiv preprint arXiv:2501.04352, 2025

    Clément Fuchs, Maxime Zanella, and Christophe De Vleeschouwer. Online gaussian test-time adaptation of vision-language models.arXiv preprint arXiv:2501.04352, 2025

    2025 arXiv

  3. [11]

    Clip-adapter: Better vision-language models with feature adapters.IJCV, 132(2), 2024

    Peng Gao, Shijie Geng, Renrui Zhang, Teli Ma, Rongyao Fang, Yongfeng Zhang, Hongsheng Li, and Yu Qiao. Clip-adapter: Better vision-language models with feature adapters.IJCV, 132(2), 2024

    2024

  4. [12]

    Dota: Distributional test-time adaptation of vision-language models.arXiv preprint arXiv:2409.19375, 2024

    Zongbo Han, Jialong Yang, Junfan Li, Qinghua Hu, Qianli Xu, Mike Zheng Shou, and Changqing Zhang. Dota: Distributional test-time adaptation of vision-language models.arXiv preprint arXiv:2409.19375, 2024

    2024

  5. [13]

    Discriminant analysis by gaussian mixtures.Journal of the Royal Statistical Society Series B: Statistical Methodology, 58(1):155–176, 1996

    Trevor Hastie and Robert Tibshirani. Discriminant analysis by gaussian mixtures.Journal of the Royal Statistical Society Series B: Statistical Methodology, 58(1):155–176, 1996

    1996

  6. [14]

    Patrick Helber, Benjamin Bischke, Andreas Dengel, and Damian Borth. EuroSAT: A Novel Dataset and Deep Learning Benchmark for Land Use and Land Cover Classification.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 12(7):2217– 2226, 2019

    2019

  7. [15]

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

    Dan Hendrycks, Steven Basart, Norman Mu, Saurav Kadavath, Frank Wang, Evan Dorundo, Rahul Desai, Tyler Zhu, Samyak Parajuli, Mike Guo, et al. The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization. InCVPR, 2021. 10

    2021

  8. [16]

    Benchmarking Neural Network Robustness to Common Corruptions and Perturbations

    Dan Hendrycks and Thomas Dietterich. Benchmarking Neural Network Robustness to Common Corruptions and Perturbations. InICLR, 2019

    2019

  9. [17]

    Natural Adversarial Examples

    Dan Hendrycks, Kevin Zhao, Steven Basart, Jacob Steinhardt, and Dawn Song. Natural Adversarial Examples. InCVPR, 2021

    2021

  10. [18]

    A class of invariant consistent tests for multivariate normality

    Norbert Henze and Bernd Zirkler. A class of invariant consistent tests for multivariate normality. Communications in statistics-Theory and Methods, 19(10):3595–3617, 1990

    1990

  11. [19]

    Test-time classifier adjustment module for model-agnostic domain generalization

    Yusuke Iwasawa and Yutaka Matsuo. Test-time classifier adjustment module for model-agnostic domain generalization. InNeurIPS, 2021

    2021

  12. [20]

    Transductive inference for text classification using support vector machines

    Thorsten Joachims. Transductive inference for text classification using support vector machines. InICML, 1999

    1999

  13. [21]

    Label propagation for zero-shot classification with vision-language models

    Yannis Kalantidis, Giorgos Tolias, et al. Label propagation for zero-shot classification with vision-language models. InCVPR, 2024

    2024

  14. [22]

    Efficient test-time adaptation of vision-language models

    Adilbek Karmanov, Dayan Guan, Shijian Lu, Abdulmotaleb El Saddik, and Eric Xing. Efficient test-time adaptation of vision-language models. InCVPR, 2024

    2024

  15. [23]

    3D Object Representations for Fine-Grained Categorization

    Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. 3D Object Representations for Fine-Grained Categorization. InCVPRW, 2013

    2013

  16. [24]

    Estimation of the precision matrix of a singular wishart distribution and its application in high-dimensional data

    Tatsuya Kubokawa and Muni S Srivastava. Estimation of the precision matrix of a singular wishart distribution and its application in high-dimensional data. 99(9):1906–1928, 2008

    1906

  17. [25]

    Ra-tta: Retrieval-augmented test-time adaptation for vision-language models

    Youngjun Lee, Doyoung Kim, Junhyeok Kang, Jihwan Bang, Hwanjun Song, and Jae-Gil Lee. Ra-tta: Retrieval-augmented test-time adaptation for vision-language models. InICLR, 2025

    2025

  18. [26]

    BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation. InICML, 2022

    2022

  19. [27]

    Align Before Fuse: Vision and Language Representation Learning with Momentum Distillation

    Junnan Li, Ramprasaath Selvaraju, Akhilesh Gotmare, Shafiq Joty, Caiming Xiong, and Steven Chu Hong Hoi. Align Before Fuse: Vision and Language Representation Learning with Momentum Distillation. InNeurIPS, 2021

    2021

  20. [28]

    Using discriminant analysis for multi-class classification: an experimental investigation.Knowledge and information systems, 10:453–472, 2006

    Tao Li, Shenghuo Zhu, and Mitsunori Ogihara. Using discriminant analysis for multi-class classification: an experimental investigation.Knowledge and information systems, 10:453–472, 2006

    2006

  21. [29]

    Text and image are mutually beneficial: Enhancing training-free few-shot classification with clip

    Yayuan Li, Jintao Guo, Lei Qi, Wenbin Li, and Yinghuan Shi. Text and image are mutually beneficial: Enhancing training-free few-shot classification with clip. InAAAI, 2025

    2025

  22. [30]

    Efficient and context-aware label propagation for zero-/few-shot training-free adaptation of vision-language model

    Yushu Li, Yongyi Su, Adam Goodge, Kui Jia, and Xun Xu. Efficient and context-aware label propagation for zero-/few-shot training-free adaptation of vision-language model. InICLR, 2025

    2025

  23. [31]

    Learning to propagate labels: Transductive propagation network for few-shot learning

    Yanbin Liu, Juho Lee, Minseop Park, Saehoon Kim, Eunho Yang, Sung Ju Hwang, and Yi Yang. Learning to propagate labels: Transductive propagation network for few-shot learning. InICLR, 2019

    2019

  24. [32]

    Swapprompt: Test-time prompt adaptation for vision-language models

    Xiaosong Ma, Jie Zhang, Song Guo, and Wenchao Xu. Swapprompt: Test-time prompt adaptation for vision-language models. InNeurIPS, 2023

    2023

  25. [33]

    Fine- Grained Visual Classification of Aircraft.arXiv preprint arXiv:1306.5151, 2013

    Subhransu Maji, Esa Rahtu, Juho Kannala, Matthew Blaschko, and Andrea Vedaldi. Fine- Grained Visual Classification of Aircraft.arXiv preprint arXiv:1306.5151, 2013

    2013 arXiv

  26. [34]

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

    Shu Manli, Nie Weili, Huang De-An, Yu Zhiding, Goldstein Tom, Anandkumar Anima, and Xiao Chaowei. Test-time prompt tuning for zero-shot generalization in vision-language models. InNeurIPS, 2022

    2022

  27. [35]

    Black-box test-time prompt tuning for vision-language models

    Fan’an Meng, Chaoran Cui, Hongjun Dai, and Shuai Gong. Black-box test-time prompt tuning for vision-language models. InAAAI, 2025. 11

    2025

  28. [36]

    A random-projection based test of gaussianity for stationary processes.Computational Statistics & Data Analysis, 75:124–141, 2014

    Alicia Nieto-Reyes, Juan Antonio Cuesta-Albertos, and Fabrice Gamboa. A random-projection based test of gaussianity for stationary processes.Computational Statistics & Data Analysis, 75:124–141, 2014

    2014

  29. [37]

    Automated Flower Classification over a Large Number of Classes

    Maria-Elena Nilsback and Andrew Zisserman. Automated Flower Classification over a Large Number of Classes. InICVGIP. IEEE, 2008

    2008

  30. [38]

    Cats and Dogs

    Omkar M Parkhi, Andrea Vedaldi, Andrew Zisserman, and CV Jawahar. Cats and Dogs. In CVPR, 2012

    2012

  31. [39]

    The matrix cookbook.Technical University of Denmark, 7(15):510, 2008

    Kaare Brandt Petersen, Michael Syskind Pedersen, et al. The matrix cookbook.Technical University of Denmark, 7(15):510, 2008

    2008

  32. [40]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. InICML, 2021

    2021

  33. [41]

    Do imagenet classifiers generalize to imagenet? InICML, 2019

    Benjamin Recht, Rebecca Roelofs, Ludwig Schmidt, and Vaishaal Shankar. Do imagenet classifiers generalize to imagenet? InICML, 2019

    2019

  34. [42]

    An extension of shapiro and wilk’s w test for normality to large samples

    J Patrick Royston. An extension of shapiro and wilk’s w test for normality to large samples. Journal of the Royal Statistical Society: Series C (Applied Statistics), 31(2):115–124, 1982

    1982

  35. [43]

    Align your prompts: Test-time prompting with distribution alignment for zero-shot generalization

    Jameel Hassan Abdul Samadh, Hanan Gani, Noor Hazim Hussein, Muhammad Uzair Khattak, Muzammal Naseer, Fahad Khan, and Salman Khan. Align your prompts: Test-time prompting with distribution alignment for zero-shot generalization. InNeurIPS, 2023

    2023

  36. [44]

    An analysis of variance test for normality.Biometrika, 52(3):591– 611, 1965

    S Shaphiro and MBJB Wilk. An analysis of variance test for normality.Biometrika, 52(3):591– 611, 1965

    1965

  37. [45]

    High-dimensional linear discriminant analysis classifier for spiked covariance model.Journal of Machine Learning Research, 21(112):1–24, 2020

    Houssem Sifaou, Abla Kammoun, and Mohamed-Slim Alouini. High-dimensional linear discriminant analysis classifier for spiked covariance model.Journal of Machine Learning Research, 21(112):1–24, 2020

    2020

  38. [46]

    UCF101: A Dataset of 101 Human Actions Classes from Videos in the Wild.arXiv preprint arXiv:1212.0402, 2012

    Khurram Soomro, Amir Roshan Zamir, and Mubarak Shah. UCF101: A Dataset of 101 Human Actions Classes from Videos in the Wild.arXiv preprint arXiv:1212.0402, 2012

    2012 arXiv

  39. [47]

    Just shift it: Test-time prototype shifting for zero-shot generalization with vision-language models

    Elaine Sui, Xiaohan Wang, and Serena Yeung-Levy. Just shift it: Test-time prototype shifting for zero-shot generalization with vision-language models. InWACV. IEEE, 2025

    2025

  40. [48]

    Sus-x: Training-free name-only transfer of vision-language models

    Vishaal Udandarao, Ankush Gupta, and Samuel Albanie. Sus-x: Training-free name-only transfer of vision-language models. InICCV, 2023

    2023

  41. [49]

    Discriminative gaussian process latent variable model for classification

    Raquel Urtasun and Trevor Darrell. Discriminative gaussian process latent variable model for classification. InICML, 2007

    2007

  42. [50]

    Tent: Fully test-time adaptation by entropy minimization

    Dequan Wang, Evan Shelhamer, Shaoteng Liu, Bruno Olshausen, and Trevor Darrell. Tent: Fully test-time adaptation by entropy minimization. InICLR, 2021

    2021

  43. [51]

    Learning Robust Global Representations by Penalizing Local Rredictive Power

    Haohan Wang, Songwei Ge, Zachary Lipton, and Eric P Xing. Learning Robust Global Representations by Penalizing Local Rredictive Power. InNeurIPS, 2019

    2019

  44. [52]

    A hard-to-beat baseline for training-free clip-based adaptation

    Zhengbo Wang, Jian Liang, Lijun Sheng, Ran He, Zilei Wang, and Tieniu Tan. A hard-to-beat baseline for training-free clip-based adaptation. InICLR, 2024

    2024

  45. [53]

    Is less more? exploring token condensation as training-free adaptation for clip

    Zixin Wang, Dong Gong, Sen Wang, Zi Huang, and Yadan Luo. Is less more? exploring token condensation as training-free adaptation for clip. InICCV, 2025

    2025

  46. [54]

    Sun Database: Large-Scale Scene Recognition from Abbey to Zoo

    Jianxiong Xiao, James Hays, Krista A Ehinger, Aude Oliva, and Antonio Torralba. Sun Database: Large-Scale Scene Recognition from Abbey to Zoo. InCVPR, 2010

    2010

  47. [55]

    Dynaprompt: Dynamic test-time prompt tuning

    Zehao Xiao, Shilin Yan, Jack Hong, Jiayin Cai, Xiaolong Jiang, Yao Hu, Jiayi Shen, Qi Wang, and Cees GM Snoek. Dynaprompt: Dynamic test-time prompt tuning. InICLR, 2025. 12

    2025

  48. [56]

    C-tpt: Calibrated test-time prompt tuning for vision-language models via text feature dispersion

    Hee Suk Yoon, Eunseop Yoon, Joshua Tian Jin Tee, Mark Hasegawa-Johnson, Yingzhen Li, and Chang D Yoo. C-tpt: Calibrated test-time prompt tuning for vision-language models via text feature dispersion. InICLR, 2024

    2024

  49. [57]

    Task residual for tuning vision- language models

    Tao Yu, Zhihe Lu, Xin Jin, Zhibo Chen, and Xinchao Wang. Task residual for tuning vision- language models. InCVPR, 2023

    2023

  50. [58]

    On the test-time zero-shot generalization of vision- language models: Do we really need prompt learning? InCVPR, 2024

    Maxime Zanella and Ismail Ben Ayed. On the test-time zero-shot generalization of vision- language models: Do we really need prompt learning? InCVPR, 2024

    2024

  51. [59]

    Realistic test-time adaptation of vision-language models

    Maxime Zanella, Clément Fuchs, Christophe De Vleeschouwer, and Ismail Ben Ayed. Realistic test-time adaptation of vision-language models. InCVPR, 2025

    2025

  52. [60]

    Boosting vision-language models with transduction

    Maxime Zanella, Benoît Gérin, and Ismail Ayed. Boosting vision-language models with transduction. InNeurIPS, 2024

    2024

  53. [61]

    Boosting vision-language models for histopathology classification: Predict all at once

    Maxime Zanella, Fereshteh Shakeri, Yunshi Huang, Houda Bahig, and Ismail Ben Ayed. Boosting vision-language models for histopathology classification: Predict all at once. InJ. Multivar . Anal., 2024

    2024

  54. [62]

    Dual prototype evolving for test-time generalization of vision-language models

    Ce Zhang, Simon Stepputtis, Katia Sycara, and Yaqi Xie. Dual prototype evolving for test-time generalization of vision-language models. InNeurIPS, 2024

    2024

  55. [63]

    Historical test-time prompt tuning for vision foundation models

    Jingyi Zhang, Jiaxing Huang, Xiaoqin Zhang, Ling Shao, and Shijian Lu. Historical test-time prompt tuning for vision foundation models. InNeurIPS, 2024

    2024

  56. [64]

    Tip-adapter: Training-free adaption of clip for few-shot classification

    Renrui Zhang, Wei Zhang, Rongyao Fang, Peng Gao, Kunchang Li, Jifeng Dai, Yu Qiao, and Hongsheng Li. Tip-adapter: Training-free adaption of clip for few-shot classification. InECCV. Springer, 2022

    2022

  57. [65]

    Boostadapter: Improving vision-language test-time adaptation via regional bootstrapping

    Taolin Zhang, Jinpeng Wang, Hang Guo, Tao Dai, Bin Chen, and Shu-Tao Xia. Boostadapter: Improving vision-language test-time adaptation via regional bootstrapping. InNeurIPS, 2024

    2024

  58. [66]

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

    Yabin Zhang, Wenjie Zhu, Hui Tang, Zhiyuan Ma, Kaiyang Zhou, and Lei Zhang. Dual memory networks: A versatile adaptation approach for vision-language models. InCVPR, 2024

    2024

  59. [67]

    Dpcore: Dynamic prompt coreset for continual test-time adaptation

    Yunbei Zhang, Akshay Mehra, Shuaicheng Niu, and Jihun Hamm. Dpcore: Dynamic prompt coreset for continual test-time adaptation. InICML, 2025

    2025

  60. [68]

    Learning with local and global consistency

    Dengyong Zhou, Olivier Bousquet, Thomas Lal, Jason Weston, and Bernhard Schölkopf. Learning with local and global consistency. InNeurIPS, 2003

    2003

  61. [69]

    Bayesian test-time adaptation for vision-language models

    Lihua Zhou, Mao Ye, Shuaifeng Li, Nianxin Li, Xiatian Zhu, Lei Deng, Hongbin Liu, and Zhen Lei. Bayesian test-time adaptation for vision-language models. InCVPR, 2025

    2025

  62. [70]

    Not all features matter: Enhancing few-shot clip with adaptive prior refinement

    Xiangyang Zhu, Renrui Zhang, Bowei He, Aojun Zhou, Dong Wang, Bin Zhao, and Peng Gao. Not all features matter: Enhancing few-shot clip with adaptive prior refinement. InICCV, 2023

    2023

  63. [71]

    Enhancing zero-shot vision models by label-free prompt distribution learning and bias correcting

    Xingyu Zhu, Beier Zhu, Yi Tan, Shuo Wang, Yanbin Hao, and Hanwang Zhang. Enhancing zero-shot vision models by label-free prompt distribution learning and bias correcting. In NeurIPS, 2024

    2024

  64. [72]

    Awt: Transferring vision-language models via augmentation, weighting, and transportation

    Yuhan Zhu, Yuyang Ji, Zhiyu Zhao, Gangshan Wu, and Limin Wang. Awt: Transferring vision-language models via augmentation, weighting, and transportation. InNeurIPS, 2024

    2024

  65. [73]

    Efficient test-time prompt tuning for vision-language models.arXiv preprint arXiv:2408.05775, 2024

    Yuhan Zhu, Guozhen Zhang, Chen Xu, Haocheng Shen, Xiaoxin Chen, Gangshan Wu, and Limin Wang. Efficient test-time prompt tuning for vision-language models.arXiv preprint arXiv:2408.05775, 2024

    2024 arXiv

  66. [74]

    Laplacian regularized few-shot learning

    Imtiaz Ziko, Jose Dolz, Eric Granger, and Ismail Ben Ayed. Laplacian regularized few-shot learning. InICML, 2020. 13 Technical Appendices and Supplementary Material This appendix provides a detailed theoretical analysis of our method, along with additional experi- mental resul...

    2020

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.