Pith. sign in

REVIEW 3 major objections 4 minor 40 references

Compositional Zero-Shot Learning with Contextualized Cues and Adaptive Contrastive Training

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read A CLIP-based framework called ULAO claims state-of-the-art compositional zero-shot learning by predicting objects before attributes and adaptively contrasting hard negatives.

desk verdict Solid, honest CZSL paper with a genuinely novel two-module design; the SOTA claim is conditional on the promised code/error bars and on how much object-cue errors propagate. read the letter →

arxiv 2412.07161 v1 pith:CKHJFFWH submitted 2024-12-10 cs.CV

classification cs.CV
keywords compositionalzero-shotlearningvision-languagemodelsCLIPsequentialprimitivepredictionhardnegativegenerationadaptivecontrastivelossattribute-objectcompositionopen-worldrecognition
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 fix a specific failure mode of CLIP-based models in compositional zero-shot learning: they confuse attributes for objects (calling a draped silk dress 'bright orange') and link attributes to the wrong object ('ruffled pants' instead of 'ruffled jacket'). The proposed ULAO framework solves this by predicting the object first, using the top-k predicted objects as weighted contextual hints to refine attribute prediction, and then training a contrastive module that generates hard negative compositions from the model's own most-confused predictions and pushes them away with an adaptively growing margin. The paper reports state-of-the-art harmonic-mean and AUC numbers across MIT-States, UT-Zappos, and C-GQA in both closed-world and open-world settings. If correct, the gain comes from modeling, not from changing the CLIP backbone.

What carries the argument

The load-bearing mechanism is the two-module pipeline. UAO first computes object probabilities from an object-specific visual attention stream using the prompt 'a photo of [object]', then replaces the generic 'object' token in the attribute prompt with a probability-weighted mixture of the top-k predicted objects' embeddings, detached from gradient flow; the composition score is the product of attribute and object probabilities. LAO adds a composition branch whose visual features are the image feature plus a learned mixer output, and a fully learnable prompt '[a] [photo] [of] [attribute] [object]'; it generates exactly two hard negatives per training image—the most probable wrong attribute paired with the true object, and the true attribute paired with the most probable wrong object, both restricted to seen compositions—and applies a contrastive loss with a per-negative threshold that starts at zero and grows with the running mean similarity gap, capped by an upper bound.

What would settle it

Take a set of images in which object recognition is artificially corrupted—crop out or mask the object regions, or replace the top-1 object with a wrong but plausible object prompt—and measure whether UAO's attribute accuracy drops more than that of a joint feature baseline; if attribute accuracy is robust to wrong object cues, the contextual-hint mechanism is not doing the claimed work. Alternatively, train UAO with k=0 (no object hints) and check whether the reported HM/AUC gap on MIT-States persists.

Watch

Extended reading notes

Core claim

The central claim is that splitting CZSL into sequential primitive prediction—object first, attribute second with object context—and then linking primitives through contrastive learning with model-generated hard negatives delivers consistent state-of-the-art results. On the three benchmarks the paper reports the best HM and AUC in both closed-world and open-world evaluation, with the largest relative gain (19.6%) on the open-world setting of C-GQA, where the search space is largest. The paper attributes the gain to two mechanisms: UAO reduces attribute-object confusion by letting object recognition proceed without attribute interference, and LAO shrinks the effective search space by penalizing exactly the mislinkages the model is most prone to.

Load-bearing premise

The whole pipeline stands on the assumption that the model's object-first predictions are accurate enough to serve as trustworthy contextual cues; the paper itself notes that 'Transitioning to UAO can lead to errors if the object cues are incorrect'.

Editorial extensions

If this is right

  • CZSL models built on frozen CLIP can be improved without fine-tuning the backbone, by changing the order and context of primitive prediction.
  • Hard negatives derived from the model's own confusions are more useful than random or composition-branch confusions for learning attribute-object linkages.
  • The adaptive margin lets the model start with easy distinctions and tighten as training proceeds, which the ablations show beats fixed or max-difference thresholds.
  • On datasets with highly similar primitives (UT-Zappos), the contrastive module contributes most; on datasets with abstract overlapping categories (MIT-States), sequential object-first prediction with object hints contributes most.

Reading between the lines

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

  • The object-first design implicitly assumes objects are more reliably recognized than attributes; in scenes with multiple objects or occlusion, the top-k object mix could average over unrelated objects and degrade attribute cues—a testable prediction: performance should drop more on multi-object images than single-object ones.
  • Because the hard negatives are restricted to seen compositions, the model never sees a negative involving an unseen object or attribute; extending generation to the open-world label space with feasibility filtering could further shrink the search space.
  • The failures the paper reports (general labels penalized toward specific ones, e.g., Sliced-Fruit -> Sliced-Apple) suggest the adaptive threshold systematically punishes high-frequency generic labels; a frequency-aware margin could counteract this.
  • The composition visual mixer is a simple additive residual on top of CLIP features; plugging the same UAO+LAO losses into stronger visual representations should compound the gains if the mechanism is the claimed modeling advantage.
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 / 4 minor

Summary. The paper proposes ULAO, a CLIP-based framework for compositional zero-shot learning (CZSL) with two modules. The UAO module performs object-first sequential prediction and uses top-k predicted object embeddings as contextual cues in attribute prompts (Eq. 3). The LAO module adds a direct composition-prediction branch and a contrastive loss whose hard negatives are generated from the UAO module's most probable attribute/object mispredictions (Eqs. 7-8), with an adaptively updated threshold (Eqs. 9-11). The authors report state-of-the-art HM and AUC results on MIT-States, UT-Zappos, and C-GQA in both closed-world and open-world settings, supported by ablations and qualitative examples.

Significance. If the reported gains reproduce, ULAO is a useful contribution to CLIP-based CZSL: it directly targets two known failure modes of CLIP, namely primitive misunderstanding and attribute-object mislinking. The idea of constructing textual hard negatives from the model's own most-confusing predictions is interesting and is tested against random and composition-branch alternatives. The paper also explicitly discusses limitations, which is commendable. The main uncertainty is empirical: the central SOTA claim rests on small margins for some datasets, and the results are reported without variance estimates, while the key design assumption about the reliability of UAO object cues is not directly tested.

major comments (3)
  1. [Comparisons with SOTAs (Table 2)] The claim that ULAO 'consistently outperforms other methods on HM and AUC across all datasets' is not yet supported with statistical evidence. No error bars, confidence intervals, or multiple-seed results are reported. On MIT-States CW, the HM margin over CDSCZSL is 0.1 point (39.3 vs. 39.2), and on MIT-States OW the unseen accuracy is actually lower than CDSCZSL (21.6 vs. 21.8). Please run at least three seeds and report mean ± standard deviation (or comparable dispersion measures) for the main comparisons, and state whether the differences are consistent across seeds.
  2. [Method (Eqs. 3, 7-8) and Limitations] Both novel modules inherit the UAO object predictions: Eq. (3) constructs the attribute-prompt object token as a probability-weighted combination of top-k predicted object embeddings, and Eqs. (7)-(8) select hard negatives by argmax over the same UAO attribute and object probabilities. The paper itself notes that 'Transitioning to UAO can lead to errors if the object cues are incorrect,' and Fig. 5 shows concrete failures (e.g., Ancient-Computer → Old-Computer, Browned-Cake → Sliced-Cake). Because several headline margins are small, the SOTA claim is conditional on the reliability of these object cues. I request an oracle-cue ablation that feeds ground-truth object labels into Eq. (3) and/or Eqs. (7)-(8), together with object-only accuracy on unseen compositions, to determine whether the gains are robust when object cues are correct.
  3. [Eqs. (7)-(8), Hard negative generation] The generated negatives must belong to the seen composition set Y^S, but the paper does not specify what happens when no seen composition (a', co) or (ca, o') exists for a given ground-truth pair. In sparse datasets such as C-GQA, such missing negatives could exclude many training examples from the contrastive loss. Please clarify whether those examples are skipped or receive a fallback, and report the coverage rate of negative generation per dataset.
minor comments (4)
  1. [Throughout] There are several typos and formatting artifacts: 'Prediciton' in Fig. 1, 'an then' in the Introduction, 'modu les' in the Final prediction and training paragraph, and 'GPICOL' vs. 'GIPCOL' in Table 2 versus the reference title.
  2. [Table 4] The row labels in Table 4 are not readable; the columns 'C', 'Random', 'UAO', 'Mean', 'Max', 'Fixed', 'up' need explicit row headers and checkmarks so the reader can tell which variant each row corresponds to.
  3. [Implementation Details] The text states that 'The provided Supplementary materials include codes, environment requirements, and detailed config files,' but no supplementary material or link is accessible in the arXiv version. Please provide an anonymized code/data link or include the config files as actual supplementary material.
  4. [Eq. (10)] The threshold update th_t = max(th_{t-1}, delta_{t-1}) makes the threshold monotonically nondecreasing up to the bound; the term 'adaptive' would be more accurate if the threshold could also decrease when the model improves. Please justify the one-sided ratchet or test a two-sided update.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ULAO is trained on seen labels and evaluated on held-out unseen compositions; self-referential hard-negative mining and adaptive thresholds are standard mechanisms, not definitional reductions.

full rationale

Walking the derivation chain, no load-bearing step reduces to its own inputs. The UAO module's object probabilities (Eq. 2) and attribute probabilities (Eq. 4) are supervised by ground-truth primitive labels through LUAO (Eq. 5) on the seen set S, and the final fused prediction p(y|x)=alpha*p(y=(ca,co)|x)+(1-alpha)*p(a|x)p(o|x) is evaluated on unseen compositions Y^U that are never used in training. The LAO module's hard negatives (Eqs. 7-8) are selected using the model's own UAO probabilities, but they are constrained to the seen composition set and contrasted against the ground-truth label (ca,co) in Eq. 9, so this is self-referential hard-negative mining rather than fitting the evaluation target. The adaptive threshold (Eq. 10) tracks the model's current margin from its own similarities, a curriculum mechanism, not a fitted prediction of unseen performance. The paper explicitly acknowledges in 'Limitations and potentials' that 'Transitioning to UAO can lead to errors if the object cues are incorrect,' but this is a robustness concern about error propagation, not circularity, because the object cues are still trained on ground-truth seen labels and the benchmarks use held-out compositions. Self-citations appear only as baselines or related work (e.g., CDSCZSL, DRANet, SAD-SP) and are not invoked to justify the framework's design choices; load-bearing design decisions cite external work (CLIP, HPL, DFSP, curriculum learning). No uniqueness theorem, ansatz-smuggling citation, or renaming of a known result is present. The central SOTA claim therefore has independent empirical content relative to its training inputs.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

The ledger lists seven hyperparameters that are tuned on the benchmark datasets rather than derived from theory. The paper introduces new network components, such as the visual attention modules and composition visual mixer, but these are trained model parts with no independent falsifiable handle. The approach assumes CLIP's frozen embeddings already contain enough attribute and object knowledge for the sequential prompts and contrastive losses to extract.

free parameters (7)
  • rk (contextual integration ratio) = not reported, studied over 0.2 to 1.0
    Controls how much the top-k object embedding replaces the fixed word 'object' in the attribute prompt (Eq. 3), tuned on benchmarks.
  • k (number of top object hints) = not reported, studied over 1 to 9
    Number of predicted objects fused into the attribute prompt (Eq. 3), affects attribute classification and is dataset-dependent.
  • rao (loss coefficient in UAO) = not reported
    Weight for composition log-probability in the UAO loss (Eq. 5).
  • rc (contrastive loss weight) = best around 0.2 to 0.4 on MIT-States (Fig. 4)
    Balances the LAO contrastive loss (Eq. 12).
  • alpha (fusion weight) = not reported
    Weight in final prediction p(y|x) = alpha * p_composition + (1-alpha) * p(a)p(o).
  • up (threshold upper bound) = not reported
    Caps the adaptive margin in Eq. (11) to prevent over-penalization.
  • rm (composition mixer weight) = not reported
    Weighting parameter in the composition visual mixer v'_c = v_c + rm * fm(v_a, v_o).
assumptions (3)
  • domain assumption CLIP's frozen visual and textual encoders provide a representational space in which attributes and objects are sufficiently separable for the task
    All modules build on CLIP ViT-L/14 embeddings; if CLIP lacks the primitive knowledge, UAO and LAO cannot recover it.
  • domain assumption Object-first sequential prediction with detached gradients prevents attribute features from corrupting object recognition and gives reliable context cues
    The authors state this design choice in the UAO section, but also acknowledge in Limitations that object cues can be wrong, which would propagate errors.
  • domain assumption The most probable mispredictions from UAO are the most informative hard negatives for learning attribute-object linkages
    LAO's negative set is built from argmax UAO confusions (Eqs. 7-8). If those confusions are not the discriminative ones, the contrastive loss may be misaligned.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Compositional Zero-Shot Learning with Contextualized Cues and Adaptive Contrastive Training." pith.science (2026). https://pith.science/paper/CKHJFFWH

@misc{pith2026241207161,
  author       = {Pith},
  title        = {Pith review of: Compositional Zero-Shot Learning with Contextualized Cues and Adaptive Contrastive Training},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CKHJFFWH}},
  note         = {Machine review of arXiv:2412.07161}
}
read the original abstract

Compositional Zero-Shot Learning (CZSL) aims to recognize unseen combinations of seen attributes and objects. Current CLIP-based methods in CZSL, despite their advancements, often fail to effectively understand and link the attributes and objects due to inherent limitations in CLIP's pretraining mechanisms. To address these shortcomings, this paper introduces a novel framework, Understanding and Linking Attributes and Objects (ULAO) in CZSL, which comprises two innovative modules. The Understanding Attributes and Objects (UAO) module improves primitive understanding by sequential primitive prediction and leveraging recognized objects as contextual hints for attribute classification. Concurrently, the Linking Attributes and Objects (LAO) module improves the attribute-object linkage understanding through a new contrastive learning strategy that incorporates tailored hard negative generation and adaptive loss adjustments. We demonstrate our model's superiority by showcasing its state-of-the-art performance across three benchmark datasets in both Closed-World (CW) and Open-World (OW) scenarios.

Figures

Figures reproduced from arXiv: 2412.07161 by the authors.

Figure 1
Figure 1. Failure cases of using CLIP-based models to solve [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Model Overview. The proposed ULAO equips UAO and LAO to understand and link attributes and objects. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Parameter study on UAO. Methods MIT-States C Random UAO Mean Max Fixed up S U HM AUC ✓ ✓ 51.2 51.7 38.7 22.1 ✓ ✓ 50.7 52.6 39.0 22.3 ✓ ✓ 50.8 52.3 39.3 22.5 ✓ ✓ 49.5 52.3 38.7 21.7 ✓ ✓ 49.8 52.3 38.9 22.0 ✓ ✓ ✓ 51.4 52.3 39.3 22.7 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Study on contrastive loss ratio rc. tency across tests. As illustrated in Figure 3a, increasing rk gen￾erally enhances model performance, confirming the advantageous effect of integrating more object hints. Conversely, Figure 3b re￾veals that adjusting the number of ob…
Figure 5
Figure 5. Figure 5: In the upper panel, contrasting C+UAO against C, we ob [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 5
Figure 5. Figure 5: Qualitative results on different variants of ULAO. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 31 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    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 word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Bao, W.; Chen, L.; Huang, H.; and Kong, Y. 2023. Prompting Language-Informed Distribution for Compositional Zero-Shot Learning. arXiv preprint arXiv:2305.14428

  4. [4]

    Doveh, S.; Arbelle, A.; Harary, S.; Herzig, R.; Kim, D.; Cascante-Bonilla, P.; Alfassy, A.; Panda, R.; Giryes, R.; Feris, R.; et al. 2024. Dense and aligned captions (dac) promote compositional reasoning in vl models. Advances in Neural Information Processing Systems, 36

  5. [5]

    Hsieh, C.-Y.; Zhang, J.; Ma, Z.; Kembhavi, A.; and Krishna, R. 2023. SugarCrepe: Fixing Hackable Benchmarks for Vision-Language Compositionality. In Thirty-Seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track

  6. [6]

    Huang, S.; Gong, B.; Feng, Y.; Zhang, M.; Lv, Y.; and Wang, D. 2024. Troika: Multi-Path Cross-Modal Traction for Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

  7. [7]

    J.; and Adelson, E

    Isola, P.; Lim, J. J.; and Adelson, E. H. 2015. Discovering states and transformations in image collections. In Proceedings of the IEEE conference on computer vision and pattern recognition, 1383--1391

  8. [8]

    Jiang, C.; and Zhang, H. 2024. Revealing the Proximate Long-Tail Distribution in Compositional Zero-Shot Learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 2498--2506

Show all 40 references
  1. [9]

    Karthik, S.; Mancini, M.; and Akata, Z. 2022. KG-SP: Knowledge Guided Simple Primitives for Open World Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 9336--9345

  2. [10]

    P.; and Ba, J

    Kingma, D. P.; and Ba, J. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980

  3. [11]

    Lake, B. M. 2014. Towards more human-like concept learning in machines: Compositionality, causality, and learning-to-learn. Ph.D. thesis, Massachusetts Institute of Technology

  4. [12]

    Li, X.; Yang, X.; Wei, K.; Deng, C.; and Yang, M. 2022. Siamese Contrastive Embedding Network for Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 9326--9335

  5. [13]

    Li, Y.; Liu, Z.; Chen, H.; and Yao, L. 2024. Context-based and Diversity-driven Specificity in Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

  6. [14]

    Li, Y.; Liu, Z.; Jha, S.; and Yao, L. 2023. Distilled reverse attention network for open-world compositional zero-shot learning. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 1782--1791

  7. [15]

    Li, Y.-L.; Xu, Y.; Mao, X.; and Lu, C. 2020. Symmetry and group in attribute-object compositions. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11316--11325

  8. [16]

    Liu, Z.; Li, Y.; Yao, L.; Chang, X.; Fang, W.; Wu, X.; and Saddik, A. E. 2023. Simple Primitives with Feasibility- and Contextuality-Dependence for Open-World Compositional Zero-shot Learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1--18

  9. [17]

    Lu, X.; Guo, S.; Liu, Z.; and Guo, J. 2023. Decomposed soft prompt guided fusion enhancing for compositional zero-shot learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 23560--23569

  10. [18]

    F.; Xian, Y.; and Akata, Z

    Mancini, M.; Naeem, M. F.; Xian, Y.; and Akata, Z. 2021. Open world compositional zero-shot learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 5222--5230

  11. [19]

    F.; Xian, Y.; and Akata, Z

    Mancini, M.; Naeem, M. F.; Xian, Y.; and Akata, Z. 2022. Learning graph embeddings for open world compositional zero-shot learning. IEEE Transactions on Pattern Analysis and Machine Intelligence

  12. [20]

    Momeni, L.; Caron, M.; Nagrani, A.; Zisserman, A.; and Schmid, C. 2023. Verbs in action: Improving verb understanding in video-language models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 15579--15591

  13. [21]

    F.; Xian, Y.; Tombari, F.; and Akata, Z

    Naeem, M. F.; Xian, Y.; Tombari, F.; and Akata, Z. 2021. Learning graph embeddings for compositional zero-shot learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 953--962

  14. [22]

    V.; Yu, P.; and Bach, S

    Nayak, N. V.; Yu, P.; and Bach, S. H. 2023. Learning to Compose Soft Prompts for Compositional Zero-Shot Learning. In International Conference on Learning Representations

  15. [23]

    Purushwalkam, S.; Nickel, M.; Gupta, A.; and Ranzato, M. 2019. Task-driven modular networks for zero-shot compositional learning. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 3593--3602

  16. [24]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning, 8748--8763. PMLR

  17. [25]

    Saini, N.; Pham, K.; and Shrivastava, A. 2022. Disentangling Visual Embeddings for Attributes and Objects. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13658--13667

  18. [26]

    Singh, H.; Zhang, P.; Wang, Q.; Wang, M.; Xiong, W.; Du, J.; and Chen, Y. 2023. Coarse-to-Fine Contrastive Learning in Image-Text-Graph Space for Improved Vision-Language Compositionality. arXiv preprint arXiv:2305.13812

  19. [27]

    Thrush, T.; Jiang, R.; Bartolo, M.; Singh, A.; Williams, A.; Kiela, D.; and Ross, C. 2022. Winoground: Probing vision and language models for visio-linguistic compositionality. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 5238--5248

  20. [28]

    Wang, H.; Yang, M.; Wei, K.; and Deng, C. 2023 a . Hierarchical Prompt Learning for Compositional Zero-Shot Recognition. In IJCAI

  21. [29]

    Wang, Q.; Liu, L.; Jing, C.; Chen, H.; Liang, G.; Wang, P.; and Shen, C. 2023 b . Learning Conditional Attributes for Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11197--11206

  22. [30]

    Wei, K.; Yang, M.; Wang, H.; Deng, C.; and Liu, X. 2019. Adversarial fine-grained composition learning for unseen attribute-object recognition. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 3741--3749

  23. [31]

    Xu, G.; Chai, J.; and Kordjamshidi, P. 2024. GIPCOL: Graph-Injected Soft Prompting for Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 5774--5783

  24. [32]

    Xu, G.; Kordjamshidi, P.; and Chai, J. Y. 2021. Zero-shot compositional concept learning. arXiv preprint arXiv:2107.05176

  25. [33]

    Xu, Z.; Wang, G.; Wong, Y.; and Kankanhalli, M. S. 2021. Relation-aware Compositional Zero-shot Learning for Attribute-Object Pair Recognition. IEEE Transactions on Multimedia

  26. [34]

    Yang, Y.; Pan, R.; Li, X.; Yang, X.; and Deng, C. 2023. Dual-stream contrastive learning for compositional zero-shot recognition. IEEE Transactions on Multimedia

  27. [35]

    Yu, A.; and Grauman, K. 2014. Fine-grained visual comparisons with local learning. In Proceedings of the IEEE conference on computer vision and pattern recognition, 192--199

  28. [36]

    Yu, A.; and Grauman, K. 2017. Semantic jitter: Dense supervision for visual comparisons via synthetic images. In Proceedings of the IEEE International Conference on Computer Vision, 5570--5579

  29. [37]

    Yuksekgonul, M.; Bianchi, F.; Kalluri, P.; Jurafsky, D.; and Zou, J. 2022. When and why vision-language models behave like bags-of-words, and what to do about it? In The Eleventh International Conference on Learning Representations

  30. [38]

    Zhang, L.; Awal, R.; and Agrawal, A. 2024. Contrasting Intra-Modal and Ranking Cross-Modal Hard Negatives to Enhance Visio-Linguistic Compositional Understanding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13774--13784

  31. [39]

    Zheng, Z.; Zhu, H.; and Nevatia, R. 2024. CAILA: Concept-Aware Intra-Layer Adapters for Compositional Zero-Shot Learning. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 1721--1731

  32. [40]

    C.; and Liu, Z

    Zhou, K.; Yang, J.; Loy, C. C.; and Liu, Z. 2022. Learning to prompt for vision-language models. International Journal of Computer Vision, 130(9): 2337--2348

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.