Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

OpenworldAUC: Towards Unified Evaluation and Optimization for Open-world Prompt Tuning

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 OpenworldAUC unifies base-to-new detection and classification into one distribution-insensitive metric for open-world prompt tuning.

desk verdict Solid metric and experiments for open-world prompt tuning, but the key equivalence in Prop. 5.1 is wrong and Thm. 5.2 needs repair. read the letter →

arxiv 2505.05180 v1 pith:PRCPQU6K submitted 2025-05-08 cs.LG

Cong Hua , Qianqian Xu , Zhiyong Yang , Zitai Wang , Shilong Bao , Qingming Huang This is my paper · ORCID
classification cs.LG
keywords OpenworldAUCopen-worldprompttuningbase-to-newdetectiongatedmixture-of-promptspairwiserankingmetricgeneralizationboundCLIPdistribution-insensitiveevaluation
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 argues that open-world prompt tuning needs one evaluation metric that rewards a model only when it both detects which domain an image belongs to and classifies the image correctly inside that domain. It shows existing metrics fall short individually and proposes OpenworldAUC, defined as the joint probability over a base/new pair that the detector ranks the base sample above the new one while both domain classifiers predict correctly. To optimize it, the paper introduces Gated Mixture-of-Prompts, a divide-and-conquer scheme with separate prompts for detection and base classification plus a fixed zero-shot new-domain classifier, and provides a generalization bound. If the proposal holds, model selection for open-world tasks becomes a single-number comparison instead of juggling HM, OverallAcc, and AUROC against each other.

What carries the argument

The central object is the OpenworldAUC pairwise metric together with its equivalent surrogate objective. The metric counts a pair $(x_b,x_n)$ only if the base sample is correctly classified, the new sample is correctly classified, and the detector ranks $x_b$ above $x_n$. Its replacement loss is $\mathbf{1}[y_b\neq g(x_b)]+\mathbf{1}[y_n\neq h(x_n)]+\mathbf{1}[y_b=g(x_b)]\mathbf{1}[r(x_b)\le r(x_n)]\mathbf{1}[y_n=h(x_n)]$, approximated by a squared ranking loss weighted by sigmoid gates derived from classifier confidence. The pseudo base-to-new partition, averaged over $K$ random splits of the known classes, simulates the unseen domain during training and produces the detector ensembles whose maximum score is used as $r$.

What would settle it

Run any competitor on an open-world split where the true new classes lie in a different distributional family from a random half of the training classes (e.g., base classes are photos and new classes are sketches or satellite images) and compare OpenworldAUC: if GMoP's OpenworldAUC collapses relative to a baseline whose detector was trained on genuine OOD signals, the pseudo-partition proxy is the bottleneck.

Watch

Extended reading notes

Core claim

OpenworldAUC, defined as $\mathbb{E}_{(x_b,y_b)\sim D_b,(x_n,y_n)\sim D_n}[\mathbf{1}[y_b=g(x_b)]\cdot \mathbf{1}[r(x_b)>r(x_n)]\cdot \mathbf{1}[y_n=h(x_n)]]$, is a unified metric for open-world prompt tuning that jointly evaluates base-to-new detection (P1), base classification (P2), and new classification (P2), while remaining insensitive to the base/new sample ratio because it uses pairwise ranking rather than class-marginal accuracy. The paper establishes that HM ignores detection, OverallAcc is dominated by the more numerous domain, and AUROC ignores classification, so none alone evaluates an open-world model faithfully. GMoP optimizes this metric by assigning a dedicated prompt to each of the detector, base classifier, and (zero-shot) new classifier, and the gating mechanism weights the pairwise ranking loss toward correctly classified samples. Theorem 5.2 bounds the generalization gap through the empirical loss, the zero-shot error of the new classifier, and three stochastic error terms involving the number of pseudo partitions, the base sample size, and the discrepancy between the pseudo and true base/new partitions. Across fifteen benchmarks, the method reports higher OpenworldAUC and better stage-wise trade-offs than ten competitors while using fewer trainable parameters than some of them.

Load-bearing premise

The whole argument assumes the pseudo base/new split of the training classes behaves like the true base/new split of the data; if that proxy is badly unrepresentative of the real new domain, the guarantee includes an unmeasured shift term and the method's new-class promise rests on the zero-shot classifier alone.

Editorial extensions

If this is right

  • Open-world model selection becomes a single-number decision guided by OpenworldAUC, replacing separate stage-wise metrics where a model strong in one stage can appear better than one that balances both.
  • Prompt tuning methods can be re-ranked by a metric that is stable under changed base/new test ratios, so deployment decisions no longer depend on assumptions about the unobserved domain distribution.
  • The gated division of labor among prompts provides a recipe for optimizing ranking-based objectives when the full label space is unavailable: pseudo partitions plus confidence gates.
  • The reported results suggest that a modest set of prompt parameters, around 26.6k, suffices to improve over methods with larger prompt budgets, which matters for low-resource adaptation of vision-language models.
  • The generalization bound implies that more pseudo partitions and more base shots monotonically reduce the excess risk, a prediction testable on any downstream benchmark.

Reading between the lines

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

  • A direct corollary: any existing prompt-tuning method could be converted into an OpenworldAUC learner solely by reweighting its training objective with the gated pairwise term, without changing the prompt architecture, so the metric itself, not GMoP, is the primary contribution.
  • The metric could be imported into other staged recognition pipelines, such as open-set recognition or test-time-agnostic long-tail learning, wherever a detector fronts a classifier and ground truth for a target split is unavailable.
  • A testable extension: swap the fixed zero-shot new classifier for a lightweight adversarially trained new-domain classifier and measure whether the irreducible zero-shot error term in Theorem 5.2 shrinks enough to improve OpenworldAUC, especially on fine-grained new classes.
  • If pseudo partitions are class-level while the true new domain is distribution-level, the bound's partition-shift term predicts a measurable degradation; comparing OpenworldAUC under class-split versus dataset-split new domains would quantify that gap and tell practitioners where the metric can be trusted.
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 proposes OpenworldAUC, a metric for open-world prompt tuning that jointly evaluates base-to-new detection and classification in both domains through a pairwise ranking formulation, and argues it is insensitive to the base/new sample ratio. To optimize the metric, it introduces GMoP, a gated mixture-of-prompts approach with a base classifier prompt, multiple detector prompts trained on pseudo base/new partitions, and a fixed zero-shot new-domain classifier. A generalization bound is stated for the empirical risk. Experiments on recognition, domain generalization, and cross-dataset benchmarks compare against ten methods and report improvements when evaluating with OpenworldAUC, with code released.

Significance. If the optimization-theory issues were repaired, the work would be a solid contribution: the metric definition (Prop. 4.2) is sound and addresses a real evaluation gap in open-world prompt tuning; the empirical study is broad (many datasets and competitors, released code); and the pseudo-partition idea for detector training is practically promising. However, the formal link between the objective actually optimized by GMoP and the OpenworldAUC metric is currently broken by an incorrect equivalence claim in Prop. 5.1, and the proof of the generalization bound contains an unsupported step that treats discrete partition distributions as continuous simplex densities. These issues are load-bearing for the central optimization and theoretical claims; once repaired, the metric and empirical method would constitute a valuable contribution. The released code supports reproducibility of the empirical results.

major comments (3)
  1. [Section 5.1, Prop. 5.1, App. A.5 and Tab. 5] The equivalence in Prop. 5.1 is incorrect. With indicators I_g=1[y_b=g(x_b)], I_r=1[r(x_b)>r(x_n)], I_h=1[y_n=h(x_n)], the arithmetic objective in Prop. 5.1 has expectation E[(1-I_g)+(1-I_h)+I_g(1-I_r)I_h] = 1 - OpenworldAUC + E[(1-I_g)(1-I_h)]. The extra term E[(1-I_g)(1-I_h)] does not vanish in general, so minimizing the stated objective is not equivalent to maximizing OpenworldAUC. The truth table in Tab. 5 is valid only if '+' is interpreted as logical OR, not arithmetic addition; the empirical risks in (OP0) and (OPfin) are arithmetic sums. Since this equivalence is the only formal link between GMoP and OpenworldAUC, the paper should either prove a surrogate-gap bound (e.g., control E[(1-I_g)(1-I_h)] under training conditions) or explicitly present the objective as a heuristic surrogate for OpenworldAUC rather than an equivalent reformulation.
  2. [Section 5.2, Thm. 5.2, App. B.4, term (a)] Term (a) of the proof of Thm. 5.2 bounds |E_Y[R_D] - E_hatY[R_hatD]| by an integral over the simplex S^{C-1} of |E(P)-E'(P)| dP, and then uses Lemma B.9 (simplex volume 1/C!) to obtain the factor ||E(P)-E'(P)||_infinity/C!. However, E and E' are, by definition, distributions over discrete class partitions, not probability densities on the continuous simplex; the integral representation and the resulting C! factor are not justified. Moreover, the term ||E(P)-E'(P)||_infinity/(C_b+C_n)! depends on the unknown true partition distribution and is never estimated or controlled by the algorithm, so the stated guarantee is conditional on a quantity the method does not handle. Please repair the derivation or restate the theorem explicitly as a conditional bound dependent on an unestimated partition-shift term.
  3. [Section 4.2, Prop. 4.3, App. A.4] The claimed lower bound on HitRaten is not proportional to OpenworldAUC u. From the derived inequality, HitRaten >= 1 - (1-u)/a where a = MissRateb; this is an affine function of u, and for a < 1-u the bound is negative and vacuous. Consequently, the subsequent claim that optimizing OpenworldAUC 'can guarantee the model performance on the new domain under arbitrary base/new ratio' (Sec. 4, discussion after Prop. 4.3) is not supported as stated. Please qualify the statement with the condition u > 1 - a and describe the bound as affine rather than proportional, or remove the overclaim.
minor comments (6)
  1. [Section 6.1] The sentence 'based on the open-source library open-source library XCurve' contains a duplicated phrase; please remove one instance.
  2. [Figure 1 caption] The caption labels the third and fourth panels both with 'c)'; the fourth panel should be labeled 'd)'.
  3. [Appendix table of contents] The word 'dasdsa' appears before the table of contents in the appendix; this appears to be an accidental insertion and should be deleted.
  4. [Section 5.2 / App. B.4] The notation for the partition-shift term is inconsistent: the theorem statement in the main text uses ||E-E'||_infinity/(C_b+C_n)!, while the proof in App. B.4 uses ||E(P)-E'(P)||_infinity/C!. Please unify the notation.
  5. [Abstract / Section 6.1] The abstract states '15 benchmarks', while the experimental section reports 11 recognition datasets, 4 ImageNet variants, 7 cross-dataset targets, and 10 imbalance variants; please clarify the exact counting of benchmarks to avoid apparent inconsistency.
  6. [Section 4.2, Prop. 4.2] The phrase 'joint possibility' should be 'joint probability' in the statement of Prop. 4.2.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: OpenworldAUC is a definition, GMoP is benchmarked against external methods, and the generalization bound is self-contained; the Prop. 5.1 algebra flaw is a correctness risk rather than a circular step.

full rationale

The paper's central claim is that OpenworldAUC, defined as E[1[y_b=g(x_b)]·1[r(x_b)>r(x_n)]·1[y_n=h(x_n)]], unifies detection and classification. This is a definition, not a derivation from a hidden input, and the paper's own proof of Prop. 4.2 computes the integral form to this joint probability without importing the conclusion. GMoP is an optimization heuristic whose objective is connected to OpenworldAUC through Prop. 5.1; the proof of that proposition treats the arithmetic '+' as Boolean OR, so the stated equivalence E[¬I_g + I_g·¬I_r·I_h + ¬I_h] = 1 - I_g I_r I_h is false as written, and the surrogate objectives (OP0)/(OPfin) are therefore not provably equivalent to maximizing OpenworldAUC. This is a genuine correctness defect, but it is not circularity: the method's effectiveness is established empirically on 15 held-out benchmarks against ten external competitors, and no parameter is fitted to the reported OpenworldAUC values. The pseudo base-to-new partition taken from DeCoOp is an acknowledged approximation: Theorem 5.2 explicitly includes the unestimated term ||E(P)-E'(P)||_inf/C! and the irreducible zero-shot error E_Dn[ell_ce(s_n,y_n)], and App. B.4 states 'Without prior knowledge of the new domain, we cannot bound the expected error.' The generalization proof uses standard McDiarmid and covering-number arguments, not the target result. Self-citations to the authors' AUC work (surrogate losses, CE regularization, OpenAUC) are auxiliary and are accompanied by external citations such as Gao & Zhou and Yuan et al., so they are not load-bearing. Appendix C.5's reformulation of OpenworldAUC as AUROC on masked scores is an equivalent computation of the same definition, not a renamed known result or a fitted prediction. Overall, the derivation chain contains no step that reduces to its own inputs by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central metric claim has no fitted constants. The method uses three hyperparameters (K, lambda, prompt length) that are tuned on validation data. The generalization guarantee depends on an unestimated distribution-mismatch term and a questionable density treatment of discrete partition distributions.

free parameters (3)
  • K (number of pseudo base-to-new partitions) = 3
    Chosen for efficiency-performance balance; sensitivity analysis in Fig.6(a) shows monotonic improvement with K, so K=3 is a hand-picked trade-off rather than a value predicted by theory.
  • lambda (cross-entropy regularization weight) = 0.5 to 1.0
    Selected from sensitivity analysis (Fig.6(c)); optimal in the range [1/2,1]. A method hyperparameter affecting the trade-off between ranking loss and CE loss.
  • Detector prompt length = 16
    Set to match DeCoOp's parameter count; not part of the metric claim.
assumptions (4)
  • domain assumption The open-world task is two-stage: first detect base vs new, then classify within the detected domain.
    Problem definition in Sec.2; the entire metric and method are built on this pipeline.
  • domain assumption Pseudo base-to-new partitions of the training classes approximate the real base/new distribution (small ||E(P)-E'(P)||infinity/C! term).
    Needed for the generalization bound in Thm.5.2; without it the theoretical guarantee is vacuous. The paper does not estimate this mismatch.
  • standard math The loss functions and prompts are Lipschitz continuous and bounded, and the covering number of the hypothesis class satisfies log N(epsilon;H,rho) <= d log(3r/epsilon).
    Assumed in App.B (Assumption 1, Lemma B.3) to apply McDiarmid and covering-number arguments.
  • ad hoc to paper In term (a) of the Thm.5.2 proof, the distributions E(P) and E'(P) over partitions are treated as densities on the probability simplex so the integral is bounded by the simplex volume 1/C!.
    This step is suspicious: discrete partition distributions are atomic measures, not absolutely continuous densities, so the stated volume bound does not follow; a total-variation bound would apply.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OpenworldAUC: Towards Unified Evaluation and Optimization for Open-world Prompt Tuning." pith.science (2026). https://pith.science/paper/PRCPQU6K

@misc{pith2026250505180,
  author       = {Pith},
  title        = {Pith review of: OpenworldAUC: Towards Unified Evaluation and Optimization for Open-world Prompt Tuning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PRCPQU6K}},
  note         = {Machine review of arXiv:2505.05180}
}
read the original abstract

Prompt tuning adapts Vision-Language Models like CLIP to open-world tasks with minimal training costs. In this direction, one typical paradigm evaluates model performance separately on known classes (i.e., base domain) and unseen classes (i.e., new domain). However, real-world scenarios require models to handle inputs without prior domain knowledge. This practical challenge has spurred the development of open-world prompt tuning, which demands a unified evaluation of two stages: 1) detecting whether an input belongs to the base or new domain (P1), and 2) classifying the sample into its correct class (P2). What's more, as domain distributions are generally unknown, a proper metric should be insensitive to varying base/new sample ratios (P3). However, we find that current metrics, including HM, overall accuracy, and AUROC, fail to satisfy these three properties simultaneously. To bridge this gap, we propose OpenworldAUC, a unified metric that jointly assesses detection and classification through pairwise instance comparisons. To optimize OpenworldAUC effectively, we introduce Gated Mixture-of-Prompts (GMoP), which employs domain-specific prompts and a gating mechanism to dynamically balance detection and classification. Theoretical guarantees ensure generalization of GMoP under practical conditions. Experiments on 15 benchmarks in open-world scenarios show GMoP achieves SOTA performance on OpenworldAUC and other metrics. We release the code at https://github.com/huacong/OpenworldAUC

Figures

Figures reproduced from arXiv: 2505.05180 by the authors.

Figure 1
Figure 1. OpenworldAUC and its optimization framework. a) We first perform pseudo partition on the training dataset to simulate new domain. b) Based on this partition, we calculate the AUROC-like ranking loss to optimize the detector. c) Then, to optimize base-domain classifier, the CE loss is calculated on the original training set. Herein, a gating mechanism selects the correctly-classified samples to calculate the aforemen… view at source ↗
Figure 2
Figure 2. The sensitive analysis of existing metric w.r.t. new/base ratio. OverallAcc is sensitive to the domain distribution, while other metrics remains stable with varying ratios of new samples. Compared with HM, OverallAcc considers detection to some extent. To be concrete, the prediction could be re￾garded as correct only if the input has been identified in the correct domain. In other words, there exists an implicit det… view at source ↗
Figure 3
Figure 3. The MissRaten-HitRateb curve on SUN397 and Flow￾ers102. Our method can outperform other competitors on the meaningful region with lower MissRateb and higher HitRaten. with a moderate Nb due to the small parameter space of the prompt. Second, the error from finite pseudo base-to-new partitions during detector optimization, which is reduced by increasing partition number K. Third, the error from a po￾tential shift fro… view at source ↗
Figures from the paper (9 more)
Figure 5
Figure 5. Figure 5: Tradeoff between per￾formance and prompt complex￾ity across different methods. (a) Effect of K (b) Effect of Shots 92.0 92.5 93.0 0 0.3 0.5 0.7 1 1.5 Lambda OpenworldAUC (c) Sensitivity Analysis of λ 48 50 52 Single W/o ZS h GMoP OpenworldAUC (d) Ablation study of Prom…
Figure 6
Figure 6. Figure 6: Sensitivity Analysis and Ablation Study. (a), (b), (c) and (d) are performed on Flowers102, DTD, Caltech101 and ImageNet, respectively. More results are provided in App.D.7, App.D.8, App.D.10 and App.D.11. form other competitors on the region with MissRateb ≤ α and Hit…
Figure 7
Figure 7. Figure 7: Trade-off between the first-stage AUROC metric and the second-stage HM metric on the Openworld Recognition Task. (a) ImageNet Sketch (b) ImageNetV2 (c) ImageNetA (d) ImageNetR [PITH_FULL_IMAGE:figures/full_fig_p040_7.png]
Figure 8
Figure 8. Figure 8: Trade-off between the first-stage AUROC metric and the second-stage HM metric on the Openworld domain generalization task. base samples remain unchanged and new samples are decreased. Under this setup, when the new/base ratio is larger than 1 meaning that there are mor…
Figure 9
Figure 9. Figure 9: Fine-grained results on four datasets, Flowers102, ImageNet, ImageNetV2, ImageNet Sketch, comparing three competitive methods and ours. Our method achieves better OpenworldAUC, indicating improved trade-offs across three sub-metrics, which further validates the compreh…
Figure 10
Figure 10. Figure 10: The Effect of Using Different Number of base-to-new Partitions (K). (a) DTD (b) OxfordPets [PITH_FULL_IMAGE:figures/full_fig_p045_10.png]
Figure 11
Figure 11. Figure 11: Comparison of few-shot learning with 1, 2, 4, 8, 16-shot samples on DTD and OxfordPets. 45 [PITH_FULL_IMAGE:figures/full_fig_p045_11.png]
Figure 12
Figure 12. Figure 12: The Sensitive Analysis of λ. 44 48 52 56 single prompt w/o ZS h GMoP (a) StanfordCars 20 25 30 35 40 single prompt w/o ZS h GMoP (b) DTD 50 55 60 65 70 75 single prompt w/o ZS h GMoP (c) Flowers102 48 50 52 single prompt w/o ZS h GMoP (d) ImageNet 45 50 55 60 single p…
Figure 13
Figure 13. Figure 13: The ablation study of mixture-of-prompts. 46 [PITH_FULL_IMAGE:figures/full_fig_p046_13.png]

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. UniTraffic-Agent: Unified Traffic Video Reasoning for AI City Challenge 2026 Track 3 with Two Out-of-Domain Evaluations

    cs.CV 2026-08 conditional novelty 4.0 of 10

    UniTraffic-Agent, a single GPT-5.5-based pipeline, ranks 2nd on FETV and 4th on PSI-VQA on the AI City Challenge 2026 public leaderboard.

Reference graph

Works this paper leans on

60 extracted references · 54 canonical work pages · cited by 1 Pith paper

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Bossard, L., Guillaumin, M., and Gool, L. V. Food-101 - mining discriminative components with random forests. In European Conference on Computer Vision, pp.\ 446--461, 2014

    work page 2014

  3. [3]

    and Tzimiropoulos, G

    Bulat, A. and Tzimiropoulos, G. LASP: Text-to-Text Optimization for Language-Aware Soft Prompting of Vision Language Models . In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 23232--23241, 2023

    work page 2023

  4. [4]

    Describing textures in the wild

    Cimpoi, M., Maji, S., Kokkinos, I., Mohamed, S., and Vedaldi, A. Describing textures in the wild. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 3606--3613, 2014

    work page 2014

  5. [5]

    Imagenet: A large-scale hierarchical image database

    Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. Imagenet: A large-scale hierarchical image database. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 248--255, 2009

    work page 2009

  6. [6]

    An introduction to ROC analysis

    Fawcett, T. An introduction to ROC analysis. Pattern Recognit. Lett., 27 0 (8): 0 861--874, 2006

    work page 2006

  7. [7]

    Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories

    Fei - Fei, L., Fergus, R., and Perona, P. Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories. In IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops , pp.\ 178--178, 2004

    work page 2004

  8. [8]

    Shrec'22 track: Open-set 3d object retrieval

    Feng, Y., Gao, Y., Zhao, X., Guo, Y., Bagewadi, N., Bui, N.-T., Dao, H., Gangisetty, S., Guan, R., Han, X., Hua, C., Hunakunti, C., Jiang, Y., Jiao, S., Ke, Y., Kuang, L., Liu, A., Nguyen, D.-H., Nguyen, H.-D., Nie, W., Pham, B.-D., Raikar, K., Tang, Q., Tran, M.-T., Wan, J., Yan, C., You, H., and Zhu, D. Shrec'22 track: Open-set 3d object retrieval. Comp...

    work page 2022

Show all 60 references
  1. [9]

    and Zhou, Z

    Gao, W. and Zhou, Z. On the consistency of auc pairwise optimization. In International Joint Conference on Artificial Intelligence, pp.\ 939--945, 2015

    2015

  2. [10]

    Aucseg: Auc-oriented pixel-level long-tail semantic segmentation

    Han, B., Xu, Q., Yang, Z., Bao, S., Wen, P., Jiang, Y., and Huang, Q. Aucseg: Auc-oriented pixel-level long-tail semantic segmentation. In Annual Conference on Neural Information Processing Systems, pp.\ 126863--126907, 2024

    2024

  3. [11]

    Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification

    Helber, P., Bischke, B., Dengel, A., and Borth, D. Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification. IEEE J. Sel. Top. Appl. Earth Obs. Remote. Sens. , 12 0 (7): 0 2217--2226, 2019

    2019

  4. [12]

    and Gimpel, K

    Hendrycks, D. and Gimpel, K. A baseline for detecting misclassified and out-of-distribution examples in neural networks. In International Conference on Learning Representations, pp.\ 1--12, 2017

    2017

  5. [13]

    The many faces of robustness: A critical analysis of out-of-distribution generalization

    Hendrycks, D., Basart, S., Mu, N., Kadavath, S., Wang, F., Dorundo, E., Desai, R., Zhu, T., Parajuli, S., Guo, M., Song, D., Steinhardt, J., and Gilmer, J. The many faces of robustness: A critical analysis of out-of-distribution generalization. In IEEE/CVF International Confer...

    2021

  6. [14]

    Natural adversarial examples

    Hendrycks, D., Zhao, K., Basart, S., Steinhardt, J., and Song, D. Natural adversarial examples. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 15262--15271, 2021 b

    2021

  7. [15]

    Reconboost: Boosting can achieve modality reconcilement

    Hua, C., Xu, Q., Bao, S., Yang, Z., and Huang, Q. Reconboost: Boosting can achieve modality reconcilement. In International Conference on Machine Learning, pp.\ 19573--19597, 2024

    2024

  8. [16]

    U., Rasheed, H

    Khattak, M. U., Rasheed, H. A., Maaz, M., Khan, S. H., and Khan, F. S. Maple: Multi-modal prompt learning. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 19113--19122, 2023 a

    2023

  9. [17]

    U., Wasim, S

    Khattak, M. U., Wasim, S. T., Naseer, M., Khan, S., Yang, M.-H., and Khan, F. S. Self-regulating prompts: Foundational model adaptation without forgetting. In IEEE/CVF International Conference on Computer Vision , pp.\ 15190--15200, 2023 b

    2023

  10. [18]

    U., Naeem, M

    Khattak, M. U., Naeem, M. F., Naseer, M., Gool, L. V., and Tombari, F. Learning to prompt with text only supervision for vision-language models. CoRR, abs/2401.02418, 2024

    2024 arXiv

  11. [19]

    and Ramanan, D

    Kong, S. and Ramanan, D. Opengan: Open-set recognition via open data generation. IEEE Trans. Pattern Anal. Mach. Intell. , 47 0 (5): 0 3233--3243, 2025

    2025

  12. [20]

    3d object representations for fine-grained categorization

    Krause, J., Stark, M., Deng, J., and Fei-Fei, L. 3d object representations for fine-grained categorization. In IEEE/CVF International Conference on Computer Vision Workshop , pp.\ 554--561, 2013

    2013

  13. [21]

    Gallop: Learning global and local prompts for vision-language models

    Lafon, M., Ramzi, E., Rambour, C., Audebert, N., and Thome, N. Gallop: Learning global and local prompts for vision-language models. In European Conference on Computer Vision, pp.\ 264--282, 2024

    2024

  14. [22]

    The power of scale for parameter-efficient prompt tuning

    Lester, B., Al - Rfou, R., and Constant, N. The power of scale for parameter-efficient prompt tuning. In Conference on Empirical Methods in Natural Language Processing, pp.\ 3045--3059, 2021

    2021

  15. [23]

    Promptkd: Unsupervised prompt distillation for vision-language models

    Li, Z., Li, X., Fu, X., Zhang, X., Wang, W., Chen, S., and Yang, J. Promptkd: Unsupervised prompt distillation for vision-language models. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 26617--26626, 2024

    2024

  16. [24]

    Stochastic auc maximization with deep neural networks

    Liu, M., Yuan, Z., Ying, Y., and Yang, T. Stochastic auc maximization with deep neural networks. In International Conference on Learning Representations, pp.\ 1--24, 2020

    2020

  17. [25]

    B., and Vedaldi, A

    Maji, S., Rahtu, E., Kannala, J., Blaschko, M. B., and Vedaldi, A. Fine-grained visual classification of aircraft. CoRR, abs/1306.5151, 2013

    2013 arXiv

  18. [26]

    Mistretta, M., Baldrati, A., Bertini, M., and Bagdanov, A. D. Improving zero-shot generalization of learned prompts via unsupervised knowledge distillation. In European Conference on Computer Vision, pp.\ 459--477, 2024

    2024

  19. [27]

    Locoop: Few-shot out-of-distribution detection via prompt learning

    Miyai, A., Yu, Q., Irie, G., and Aizawa, K. Locoop: Few-shot out-of-distribution detection via prompt learning. In Annual Conference on Neural Information Processing Systems, pp.\ 76298--76310, 2023

    2023

  20. [28]

    Out-of-distribution detection with negative prompts

    Nie, J., Zhang, Y., Fang, Z., Liu, T., Han, B., and Tian, X. Out-of-distribution detection with negative prompts. In International Conference on Learning Representations, pp.\ 1--20, 2024

    2024

  21. [29]

    and Zisserman, A

    Nilsback, M.-E. and Zisserman, A. Automated flower classification over a large number of classes. In Indian Conference on Computer Vision, Graphics and Image Processing, pp.\ 722--729, 2008

    2008

  22. [30]

    M., Vedaldi, A., Zisserman, A., and Jawahar, C

    Parkhi, O. M., Vedaldi, A., Zisserman, A., and Jawahar, C. V. Cats and dogs. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 3498--3505, 2012

    2012

  23. [31]

    Automatic differentiation in pytorch

    Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., and Lerer, A. Automatic differentiation in pytorch. 2017

    2017

  24. [32]

    W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., and Sutskever, I

    Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., and Sutskever, I. Learning transferable visual models from natural language supervision. In International Conference on Machine Learning, pp.\ 8...

    2021

  25. [33]

    Do I mage N et classifiers generalize to I mage N et? In International Conference on Machine Learning, pp.\ 5389--5400, 2019

    Recht, B., Roelofs, R., Schmidt, L., and Shankar, V. Do I mage N et classifiers generalize to I mage N et? In International Conference on Machine Learning, pp.\ 5389--5400, 2019

    2019

  26. [34]

    and Etemad, A

    Roy, S. and Etemad, A. Consistency-guided prompt learning for vision-language models. In International Conference on Learning Representations, pp.\ 1--14, 2024

    2024

  27. [35]

    Scherreik, M. D. and Rigling, B. D. Open set recognition for automatic target classification with rejection. IEEE Trans. Aerosp. Electron. Syst. , 52 0 (2): 0 632--642, 2016

    2016

  28. [36]

    R., and Shah, M

    Soomro, K., Zamir, A. R., and Shah, M. UCF101: A dataset of 101 human actions classes from videos in the wild. CoRR, abs/1212.0402, 2012

    2012 arXiv

  29. [37]

    Kill two birds with one stone: Rethinking data augmentation for deep long-tailed learning

    Wang, B., Wang, P., Xu, W., Wang, X., Zhang, Y., Wang, K., and Wang, Y. Kill two birds with one stone: Rethinking data augmentation for deep long-tailed learning. In International Conference on Learning Representations, pp.\ 1--24, 2024 a

    2024

  30. [38]

    C., and Xing, E

    Wang, H., Ge, S., Lipton, Z. C., and Xing, E. P. Learning robust global representations by penalizing local predictive power. In Annual Conference on Neural Information Processing Systems, pp.\ 10506--10518, 2019

    2019

  31. [39]

    Llm-autoda: Large language model-driven automatic data augmentation for long-tailed problems

    Wang, P., Zhao, Z., Wen, H., Wang, F., Wang, B., Zhang, Q., and Wang, Y. Llm-autoda: Large language model-driven automatic data augmentation for long-tailed problems. In Annual Conference on Neural Information Processing Systems, pp.\ 64915--64941, 2024 b

    2024

  32. [40]

    Openauc: Towards auc-oriented open-set recognition

    Wang, Z., Xu, Q., Yang, Z., He, Y., Cao, X., and Huang, Q. Openauc: Towards auc-oriented open-set recognition. In Annual Conference on Neural Information Processing Systems, pp.\ 25033--25045, 2022

    2022

  33. [41]

    Cascade prompt learning for vision-language model adaptation

    Wu, G., Zhang, X., Li, Z., Chen, Z., Liang, J., Yang, J., and Li, X. Cascade prompt learning for vision-language model adaptation. In European Conference on Computer Vision, pp.\ 304--321, 2024

    2024

  34. [42]

    Zero-shot learning — the good, the bad and the ugly

    Xian, Y., Schiele, B., and Akata, Z. Zero-shot learning — the good, the bad and the ugly. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 3077--3086, 2017

    2017

  35. [43]

    A., Oliva, A., and Torralba, A

    Xiao, J., Hays, J., Ehinger, K. A., Oliva, A., and Torralba, A. Sun database: Large-scale scene recognition from abbey to zoo. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 3485--3492, 2010

    2010

  36. [44]

    and Ying, Y

    Yang, T. and Ying, Y. AUC maximization in the era of big data and AI: A survey. ACM Comput. Surv. , 55 0 (8): 0 172:1--172:37, 2023

    2023

  37. [45]

    Learning with multiclass AUC: theory and algorithms, 2022

    Yang, Z., Xu, Q., Bao, S., Cao, X., and Huang, Q. Learning with multiclass AUC: theory and algorithms, 2022

    2022

  38. [46]

    Optimizing two-way partial AUC with an end-to-end framework

    Yang, Z., Xu, Q., Bao, S., He, Y., Cao, X., and Huang, Q. Optimizing two-way partial AUC with an end-to-end framework. IEEE Trans. Pattern Anal. Mach. Intell. , 45 0 (8): 0 10228--10246, 2023

    2023

  39. [47]

    Harnessing hierarchical label distribution variations in test agnostic long-tail recognition

    Yang, Z., Xu, Q., Wang, Z., Li, S., Han, B., Bao, S., Cao, X., and Huang, Q. Harnessing hierarchical label distribution variations in test agnostic long-tail recognition. In International Conference on Machine Learning, pp.\ 56624--56664, 2024

    2024

  40. [48]

    Visual-language prompt tuning with knowledge-guided context optimization

    Yao, H., Zhang, R., and Xu, C. Visual-language prompt tuning with knowledge-guided context optimization. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 6757--6767, 2023

    2023

  41. [49]

    Tcp: Textual-based class-aware prompt tuning for visual-language model

    Yao, H., Zhang, R., and Xu, C. Tcp: Textual-based class-aware prompt tuning for visual-language model. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 23438--23448, 2024

    2024

  42. [50]

    Compositional training for end-to-end deep AUC maximization

    Yuan, Z., Guo, Z., Chawla, N., and Yang, T. Compositional training for end-to-end deep AUC maximization. In International Conference on Learning Representations, pp.\ 1--26, 2022

    2022

  43. [51]

    Counterfactual zero-shot and open-set visual recognition

    Yue, Z., Wang, T., Zhang, H., Sun, Q., and Hua, X.-S. Counterfactual zero-shot and open-set visual recognition. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 15404--15414, 2021

    2021

  44. [52]

    Enhancing outlier knowledge for few-shot out-of-distribution detection with extensible local prompts

    Zeng, F., Cheng, Z., Zhu, F., and Zhang, X. Enhancing outlier knowledge for few-shot out-of-distribution detection with extensible local prompts. CoRR, abs/2409.04796, 2024

    2024 arXiv

  45. [53]

    T., and Song, J

    Zhang, J., Wu, S., Gao, L., Shen, H. T., and Song, J. Dept: Decoupled prompt tuning. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 12924--12933, 2024

    2024

  46. [54]

    Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners

    Zhang, R., Hu, X., Li, B., Huang, S., Deng, H., Qiao, Y., Gao, P., and Li, H. Prompt, generate, then cache: Cascade of foundation models makes strong few-shot learners. pp.\ 15211--15222, 2023

    2023

  47. [55]

    Two fists, one heart: Multi-objective optimization based strategy fusion for long-tailed learning

    Zhao, Z., Wang, P., Wen, H., Xu, W., Song, L., Zhang, Q., and Wang, Y. Two fists, one heart: Multi-objective optimization based strategy fusion for long-tailed learning. In International Conference on Machine Learning, pp.\ 61040--61071, 2024 a

    2024

  48. [56]

    Breaking long-tailed learning bottlenecks: A controllable paradigm with hypernetwork-generated diverse experts

    Zhao, Z., Wen, H., Wang, Z., Wang, P., Wang, F., Lai, S., Zhang, Q., and Wang, Y. Breaking long-tailed learning bottlenecks: A controllable paradigm with hypernetwork-generated diverse experts. In Annual Conference on Neural Information Processing Systems, pp.\ 7493--7520, 2024 b

    2024

  49. [57]

    C., and Liu, Z

    Zhou, K., Yang, J., Loy, C. C., and Liu, Z. Conditional prompt learning for vision-language models. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp.\ 16795--16804, 2022 a

    2022

  50. [58]

    C., and Liu, Z

    Zhou, K., Yang, J., Loy, C. C., and Liu, Z. Learning to prompt for vision-language models. Int. J. Comput. Vis., 130 0 (9): 0 2337--2348, 2022 b

    2022

  51. [59]

    Decoop: Robust prompt tuning with out-of-distribution detection

    Zhou, Z., Yang, M., Shi, J.-X., Guo, L.-Z., and Li, Y.-F. Decoop: Robust prompt tuning with out-of-distribution detection. In International Conference on Machine Learning, pp.\ 62161--62177, 2024

    2024

  52. [60]

    Prompt-aligned gradient for prompt tuning

    Zhu, B., Niu, Y., Han, Y., Wu, Y., and Zhang, H. Prompt-aligned gradient for prompt tuning. In IEEE/CVF International Conference on Computer Vision , pp.\ 15613--15623, 2023

    2023

Pith tools

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