Pith. sign in

REVIEW 3 major objections 5 minor 38 references

Mining Contextualized Visual Associations from Images for Creativity Understanding

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Fine-tuning a visual encoder on mined abstract captions improves zero-shot retrieval in poetry and metaphor tasks.

desk verdict A genuinely useful dataset and a clean abstraction-ladder result, but the retrieval gains are confounded: no ablation isolates contextualized associations from synthetic captions or prefix-tuning. read the letter →

arxiv 2507.18915 v1 pith:A4SDJKYG submitted 2025-07-25 cs.CL cs.CV

classification cs.CLcs.CV
keywords contextualizedassociationscreativecaptionsabstractionladderzero-shotretrievalCLIPfine-tuningpoetry-to-imagevisualmetaphorsynthetic
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces a scalable method for mining contextualized associations—concepts that a salient object evokes in its specific scene context—and uses them to generate 1.671 million creative captions for images in MSCOCO at five degrees of abstraction. The central claim is that training a visual encoder on these captions sharpens its ability to match images with poems and metaphors, two tasks where standard alt-text-trained CLIP models underperform. The authors find statistically significant gains in zero-shot poetry-to-image retrieval, metaphor-to-visualization matching, and visual-metaphor-to-text retrieval after fine-tuning with per-abstraction prefix embeddings. Human evaluation supports that the captions stay visually grounded while exhibiting recognizably increasing abstraction. If correct, the method gives creative vision-language tasks a simple, scalable source of training signal without hand-curated datasets.

What carries the argument

The load-bearing object is the abstraction-indexed visual association, generated along a five-rung ladder: near synonyms, slightly broader categories, broader situational or emotional context, conceptual themes, and full metaphors. For each salient word, the pipeline filters MSCOCO short captions by part of speech and a concreteness lexicon, conditions a text-only LLM on a detailed VLM-written caption so the association reflects scene context rather than the word alone, and then asks a VLM to produce a sub-10-word creative caption that must include the association word. Fine-tuning adds five trainable prefix embeddings to a frozen CLIP visual encoder, one per degree, optimized with the contrastive image-text loss; at inference the prefix selects how abstractly the model should match an image to text.

What would settle it

Retrain the fine-tuned model on the same 1.671 million captions with each association word replaced by a randomly drawn association for the same word at the same abstraction degree; if the poetry-to-image and metaphor retrieval gains persist unchanged, then image-specific contextualization is not what drives the improvement, and the paper's central explanation would collapse.

Watch

Extended reading notes

Core claim

The paper claims that a three-step pipeline can turn any image with a short literal caption into a set of creative captions at controlled abstraction levels, and that these synthetic captions carry useful learning signal for creative vision-language understanding. First, a vision-language model writes a detailed caption; then a text-only LLM, conditioned on that detailed caption, mines associations for each salient word on a five-rung abstraction ladder from near-synonym to full metaphor; finally, a vision-language model rewrites the short caption into a grounded creative caption that includes the association. Fine-tuning a frozen CLIP visual encoder on this corpus, with a learnable prefix for each abstraction degree, improves zero-shot retrieval on two creative benchmarks compared with the unfine-tuned baseline, with the largest poetry retrieval gains at the highest abstraction degrees. The paper's own results also show that unfine-tuned CLIP prefers literal captions over abstract ones roughly as often as it prefers literal captions over hallucinated ones, which motivates why the mined associations are needed.

Load-bearing premise

The load-bearing premise is that a text-only LLM, given a machine-written detailed caption, mines associations that are genuinely visually grounded for the specific image; if the detailed caption contains an error or the association is generic, every downstream creative caption and the fine-tuning signal inherits that flaw.

Editorial extensions

If this is right

  • Poetry-to-image retrieval improves at abstraction degrees 4 and 5, with average rank of the correct image falling from 70.46 to 67.60 and recall@1 rising from 0.1505 to 0.1624.
  • The fine-tuned encoder prefers the correct AI-generated visualization over its literal counterpart at every abstraction degree, with near-synonym captions improving the preference rate from 0.43 to 0.59.
  • Since the generation pipeline only needs an image and a short literal caption, it can be applied to any unlabeled corpus, extending the approach to art images, social media photos, or other domains.
  • The per-degree prefix design gives practitioners a dial: the same model can be used in a literal mode or an abstract mode depending on the creative task.
  • The synthetic captions expose a weakness of literalism in standard CLIP, which at the highest abstraction degree cannot distinguish between hallucinated captions and abstract ones better than random.

Reading between the lines

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

  • One implication the paper leaves implicit: the same pipeline could mine culturally specific associations by prompting the LLM in different languages, since the association step is text-only and the authors note the current corpus is English-centric and Western-centric.
  • A natural ablation would replace the detailed VLM caption with the short caption when mining associations; if scene context is doing the real work, association quality and downstream retrieval should drop.
  • The grounding evaluation relies on the same VLM for detailed captions and creative captions, so shared error patterns could in principle inflate the grounding scores; using different VLMs for the two steps would be a sturdier test.
  • The varying optimal abstraction degree across tasks suggests a future model that predicts the right abstraction level per query, rather than using a single setting.
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 / 5 minor

Summary. The paper proposes a scalable pipeline that, for any image with a short literal caption, mines contextualized visual associations for salient elements at five degrees of abstraction (using Molmo for detailed captions and GPT-4o-mini for association mining) and then uses those associations to generate creative captions that remain visually grounded. The authors construct a dataset of about 1.7M creative captions for MSCOCO, validate the caption quality through human annotation of grounding and abstraction ranking, and fine-tune an OpenCLIP ViT-B/32 encoder with degree-specific learnable prefixes on the synthetic captions. They evaluate the fine-tuned models on three zero-shot creative retrieval tasks: poetry-to-image retrieval on MultiM-Poem, visual-metaphor-to-linguistic-metaphor retrieval on HAIVMet, and linguistic-metaphor-to-visualization matching on HAIVMet. The reported results show improved retrieval at higher abstraction degrees (d=4 and d=5) on Task 1, improved average rank on Task 2, and mixed but mostly improved preference on Task 3.

Significance. If the central claim is fully established, the work would make a useful contribution: it provides a scalable method for generating creative, increasingly abstract captions and demonstrates that training on such captions can improve zero-shot performance in creative vision-language retrieval, a domain where standard CLIP models are known to struggle. The paper has notable strengths: the dataset, generation code, and models are released; the evaluation uses external benchmarks (MultiM-Poem and HAIVMet) that were not used to fit model parameters; fine-tuning is parameter-efficient (only prefix embeddings are updated); and the human evaluation directly assesses both grounding and perceived abstraction. The paper also explicitly acknowledges limitations about hallucination and bias in the underlying LLMs/VLMs. However, the causal attribution of the retrieval improvements to contextualized associations is not yet secured, because the fine-tuned model differs from the baseline along several confounded dimensions and the human grounding evidence is based on a small, post-hoc thresholded annotation study.

major comments (3)
  1. [§5.4, Tables 4-6] The central claim of the paper is that contextualized associations, rather than synthetic captions or prefix-tuning alone, drive the observed retrieval improvements. The current comparison between OpenCLIP and OpenCLIP-FT(d) changes at least three factors simultaneously: a learned per-degree prefix, the use of 1.7M synthetic captions as training data, and the replacement of literal salient words with LLM-mined association words. Without control experiments that train the same prefix on the same images with (a) the original salient words instead of associations, (b) associations randomly permuted across images, or (c) generic non-associative synthetic captions, the gains at d=4 and d=5 cannot be attributed to contextualization. These controls are necessary to support the causal language in the abstract and in §5.4.
  2. [§5.2, Table 2] The grounding threshold (ratings 3 and 4 treated as 'visually grounded') was chosen after inspecting score-3 captions and using Gemini 2.0 Flash to argue that they contain no actual errors. This post-hoc thresholding can inflate the reported grounding percentages, and the supporting annotation is small (100 captions with a Fleiss kappa of 0.303). The claim that the captions 'remain visually grounded' is important because the training signal inherits caption quality. The authors should report the full distribution of raw ratings, the number of captions at each rating, and ideally a pre-registered or independently motivated threshold, or alternatively verify that the main retrieval results are robust to restricting training to captions with higher raw grounding scores.
  3. [§5.4, Table 6] In Task 3, the baseline OpenCLIP performance is 0.43, below random chance, and only the improvements at d=1 and d=3 are marked as statistically significant. The sentence 'All abstraction settings improve over the baseline' is misleading because d=2, d=4, and d=5 show non-significant improvements. The claim of 'meaningful improvements' across creative tasks should be tempered, or additional evidence (e.g., larger evaluation sets, confidence intervals, or multiple seeds) should be provided to establish that the Task 3 finding is reliable.
minor comments (5)
  1. [§4.2] The text first says annotators rank six captions for each of 100 images, but then states that three annotations were collected for 20% of these tasks (20 images). Please clarify the exact number of annotation tasks, images, and captions used for the abstraction ranking evaluation.
  2. [§A.1] The appendix states that the model was fine-tuned for a single epoch but also mentions early stopping with patience 3 on the validation loss. These statements are inconsistent; please clarify whether early stopping was used and what the final epoch count actually was.
  3. [§A.3.1] The appendix uses 'wilcoxin tests' and 'hyperparemters' in a few places; these should be corrected to 'Wilcoxon' and 'hyperparameters'.
  4. [§4.1 and §7] Minor typos include 'MSCOCOvalidation' (missing space) and 'gpt4o-mini' (missing hyphen). These do not affect the substance of the paper.
  5. [§5.1, Table 1] The uniqueness statistic in Table 1 would be more informative if compared to a baseline such as randomly sampled associations or associations mined from a generic (non-contextual) prompt, since even non-contextual associations may be mostly unique per image by chance.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: retrieval gains are measured on external creative-domain benchmarks, and the dataset is generated before fine-tuning; no prediction reduces to a fitted value.

full rationale

The paper's derivation chain is: (1) Molmo generates a detailed caption; (2) GPT-4o-mini mines contextualized associations at five abstraction degrees using that caption plus extracted salient words; (3) Molmo writes synthetic creative captions containing those associations; (4) a CLIP encoder is prefix-tuned on the resulting 1.7M-caption corpus; (5) the tuned model is evaluated zero-shot on MultiM-Poem and HAIVMet. No step in this chain defines an evaluation outcome in terms of the training data or a fitted parameter. The reported retrieval improvements are computed against external benchmarks that were not used to fit the prefixes, the association prompts, or the caption generator. The paper does not fit a parameter to a subset of the evaluation data and then call the fitted value a prediction; the fine-tuning corpus is constructed before the evaluation, and the evaluation tasks come from separate published datasets. The only self-citation (Ananthram et al., 2025, in the future-work section) is a pointer toward culturally specific regularities and is not load-bearing for the paper's central claims. The human abstraction-ranking study uses the same five-level ladder that was given to the generators, so it mainly validates prompt adherence rather than an independent construct, but this is not a circular derivation of the retrieval results: the headline claim about creative-domain retrieval does not depend on that ranking. Concerns raised by a skeptical reader, such as the absence of an ablation isolating contextualized associations from synthetic captions or prefix-tuning, and the post-hoc grounding threshold, are experimental-validity and attribution weaknesses, not circular reductions under the standard required here: no equation in the paper reduces to its own input, and no fitted input is renamed as a prediction. Accordingly, the paper is self-contained against external benchmarks and shows no significant circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim is supported by generated data and external benchmark evaluation, not by a formal derivation. The main burden is empirical: two hand-chosen thresholds (concreteness 3, grounding >= 3) and a set of unverified assumptions about LLM reliability and transfer. No new physical or theoretical entities are introduced.

free parameters (3)
  • Concreteness threshold for salient element extraction = 3
    Hand-chosen cutoff on Brysbaert et al. concreteness ratings that decides which nouns, adjectives, and verbs become salient elements (Section 4.1). Affects all downstream associations and captions.
  • Visual grounding acceptability threshold = score >= 3 on 4-point Likert scale
    Post hoc bucketing in Section 5.2: after analyzing score-3 captions, authors classify them as grounded; this choice increases reported grounding percentages.
  • Number of mined associations per abstraction degree = 3 words per degree per salient word
    Prompt design decision in Section A.2.2 that controls dataset size and diversity; not justified empirically.
assumptions (5)
  • domain assumption Text-only LLMs are high-quality repositories of common associations when conditioned on scene context
    Stated in Section 3.1 as justification for using GPT-4o-mini to mine contextualized associations; no independent evidence is given that these associations match human visual associations.
  • domain assumption Molmo's detailed captions are accurate enough that associations mined from them stay visually grounded
    The pipeline depends on the VLM's transcription of image content before association mining; errors propagate to captions. The paper's human evaluation is the only check and it shows only fair agreement.
  • domain assumption The prompt instruction 'semantically correct and grounded in this image' is sufficient to prevent hallucination in generated captions
    Used in Section A.2.3; the paper's own analysis shows annotators sometimes penalize interpretive language, indicating the instruction does not guarantee literal correctness.
  • domain assumption Fine-tuning a prefix on synthetic captions transfers to zero-shot creative retrieval through CLIP's shared embedding space
    No control conditions isolate this transfer; it is the experimental hypothesis rather than an established fact.
  • standard math Paired Wilcoxon and paired t-tests are appropriate for the reported differences
    Used for significance in Section A.3.1; assumes paired observations, but multiple comparisons across five degrees and three tasks are not corrected.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mining Contextualized Visual Associations from Images for Creativity Understanding." pith.science (2026). https://pith.science/paper/A4SDJKYG

@misc{pith2026250718915,
  author       = {Pith},
  title        = {Pith review of: Mining Contextualized Visual Associations from Images for Creativity Understanding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A4SDJKYG}},
  note         = {Machine review of arXiv:2507.18915}
}
read the original abstract

Understanding another person's creative output requires a shared language of association. However, when training vision-language models such as CLIP, we rely on web-scraped datasets containing short, predominantly literal, alt-text. In this work, we introduce a method for mining contextualized associations for salient visual elements in an image that can scale to any unlabeled dataset. Given an image, we can use these mined associations to generate high quality creative captions at increasing degrees of abstraction. With our method, we produce a new dataset of visual associations and 1.7m creative captions for the images in MSCOCO. Human evaluation confirms that these captions remain visually grounded while exhibiting recognizably increasing abstraction. Moreover, fine-tuning a visual encoder on this dataset yields meaningful improvements in zero-shot image-text retrieval in two creative domains: poetry and metaphor visualization. We release our dataset, our generation code and our models for use by the broader community.

Figures

Figures reproduced from arXiv: 2507.18915 by the authors.

Figure 1
Figure 1. Two images depicting trees in different set￾tings. Their alt-text makes no mention of the diverse concepts that each tree evokes. Using our method, we are able to mine contextualized associations at degrees of abstraction that extend beyond literal description. their pre-training data (Menon et al., 2024). This is especially true in creative domains. In poetry and metaphor visualization, CLIP’s capabilities are limi… view at source ↗
Figure 2
Figure 2. Our method for mining contextualized associations and generating creative captions with increasing [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Examples from our corpus. For each word, we depict its contextualized associations at increasing degrees [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Examples from each of our three evaluation tasks. Correct answers are highlighted in green. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Plots comparing OpenCLIP’s scores for original (left) and hallucinatory (right) captions against its scores [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 19 canonical work pages

  1. [1]

    Amith Ananthram, Elias Stengel-Eskin, Mohit Bansal, and Kathleen McKeown. 2025. See it from my perspective: How language affects cultural bias in image understanding. In The Thirteenth International Conference on Learning Representations

  2. [2]

    Roger E Beaty and Yoed N Kenett. 2023. Associative thinking at the core of creativity. Trends in cognitive sciences, 27(7):671--683

  3. [3]

    Tolga Bolukbasi, Kai-Wei Chang, James Y Zou, Venkatesh Saligrama, and Adam T Kalai. 2016. Man is to computer programmer as woman is to homemaker? debiasing word embeddings. Advances in neural information processing systems, 29

  4. [4]

    Florian Bordes, Richard Yuanzhe Pang, Anurag Ajay, Alexander C Li, Adrien Bardes, Suzanne Petryk, Oscar Ma \ n as, Zhiqiu Lin, Anas Mahmoud, Bargav Jayaraman, et al. 2024. An introduction to vision-language modeling. arXiv preprint arXiv:2405.17247

  5. [5]

    Marc Brysbaert, Amy Beth Warriner, and Victor Kuperman. 2014. Concreteness ratings for 40 thousand generally known english word lemmas. Behavior research methods, 46:904--911

  6. [6]

    Tuhin Chakrabarty, Vishakh Padmakumar, He He, and Nanyun Peng. 2023 a . Creative natural language generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Tutorial Abstracts, pages 34--40

  7. [7]

    Tuhin Chakrabarty, Arkadiy Saakyan, Debanjan Ghosh, and Smaranda Muresan. 2022. Flute: Figurative language understanding through textual explanations. arXiv preprint arXiv:2205.12404

  8. [8]

    Tuhin Chakrabarty, Arkadiy Saakyan, Olivia Winn, Artemis Panagopoulou, Yue Yang, Marianna Apidianaki, and Smaranda Muresan. 2023 b . I spy a metaphor: Large language models and diffusion models co-create visual metaphors. arXiv preprint arXiv:2305.14724

Show all 38 references
  1. [9]

    Lin Chen, Jinsong Li, Xiaoyi Dong, Pan Zhang, Conghui He, Jiaqi Wang, Feng Zhao, and Dahua Lin. 2024. Sharegpt4v: Improving large multi-modal models with better captions. In European Conference on Computer Vision, pages 370--387. Springer

  2. [10]

    Matt Deitke, Christopher Clark, Sangho Lee, Rohun Tripathi, Yue Yang, Jae Sung Park, Mohammadreza Salehi, Niklas Muennighoff, Kyle Lo, Luca Soldaini, et al. 2024. Molmo and pixmo: Open weights and open data for state-of-the-art multimodal models. arXiv preprint arXiv:2409.17146

  3. [11]

    Improving clip training with language rewrites

    Lijie Fan, Dilip Krishnan, Phillip Isola, Dina Katabi, and Yonglong Tian. Improving clip training with language rewrites

  4. [12]

    Joseph L Fleiss. 1971. Measuring nominal scale agreement among many raters. Psychological bulletin, 76(5):378

  5. [13]

    Ernst Hans Gombrich. 2023. Art and illusion: A study in the psychology of pictorial representation-millennium edition

  6. [14]

    Varvara Guljajeva, Mar Canet Sol \`a , and Isaac Joseph Clarke. 2023. Explaining clip through co-creative drawings and interaction. arXiv preprint arXiv:2306.07429

  7. [15]

    Samuel Ichiy \'e Hayakawa. 1967. Language in thought and action. The Florida English Journal, 3(2):1--12

  8. [16]

    Junjie Hu, Yu Cheng, Zhe Gan, Jingjing Liu, Jianfeng Gao, and Graham Neubig. 2020. What makes a good story? designing composite rewards for visual storytelling. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 7969--7976

  9. [17]

    Zheng Hu, Jiao Luo, Chunhong Zhang, and Wei Li. 2019. A natural language process-based framework for automatic association word extraction. IEEE Access, 8:1986--1997

  10. [18]

    Ting-Hao Huang, Francis Ferraro, Nasrin Mostafazadeh, Ishan Misra, Aishwarya Agrawal, Jacob Devlin, Ross Girshick, Xiaodong He, Pushmeet Kohli, Dhruv Batra, et al. 2016. Visual storytelling. In Proceedings of the 2016 conference of the North American chapter of the association...

  11. [19]

    Fanjie Kong, Yanbei Chen, Jiarui Cai, and Davide Modolo. 2024. Hyperbolic learning with synthetic captions for open-world detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16762--16771

  12. [20]

    Zhengfeng Lai, Haotian Zhang, Bowen Zhang, Wentao Wu, Haoping Bai, Aleksei Timofeev, Xianzhi Du, Zhe Gan, Jiulong Shan, Chen-Nee Chuah, et al. 2024. Veclip: Improving clip training via visual-enriched captions. In European Conference on Computer Vision, pages 111--127. Springer

  13. [21]

    Xiang Lisa Li and Percy Liang. 2021. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190

  14. [22]

    Lawrence Zitnick, and Piotr Dollár

    Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro Perona, Deva Ramanan, C. Lawrence Zitnick, and Piotr Dollár. 2014. https://doi.org/10.48550/ARXIV.1405.0312 Microsoft coco: Common objects in context . arXiv preprint

  15. [23]

    Bei Liu, Jianlong Fu, Makoto P Kato, and Masatoshi Yoshikawa. 2018. Beyond narrative description: Generating poetry from images by multi-adversarial training. In Proceedings of the 26th ACM international conference on Multimedia, pages 783--791

  16. [24]

    Yanqing Liu, Xianhang Li, Zeyu Wang, Bingchen Zhao, and Cihang Xie. 2024. Clips: An enhanced clip framework for learning with synthetic captions. arXiv preprint arXiv:2411.16828

  17. [25]

    Yue Lu, Chao Guo, Xingyuan Dai, and Fei-Yue Wang. 2022. Artcap: A dataset for image captioning of fine art paintings. IEEE Transactions on Computational Social Systems, 11(1):576--587

  18. [26]

    Sachit Menon, Ishaan Preetam Chandratreya, and Carl Vondrick. 2024. Task bias in contrastive vision-language models. International Journal of Computer Vision, 132(6):2026--2040

  19. [27]

    Saif Mohammad. 2013. Colourful language: Measuring word-colour associations. arXiv preprint arXiv:1309.5942

  20. [28]

    Saif M Mohammad and Peter D Turney. 2013. Crowdsourcing a word--emotion association lexicon. Computational intelligence, 29(3):436--465

  21. [29]

    Shibani Santurkar, Yann Dubois, Rohan Taori, Percy Liang, and Tatsunori Hashimoto. 2022. Is a caption worth a thousand images. A Controlled Study for Representation Learning. CoRR abs/2207.07635, 5

  22. [30]

    Sahand Sharifzadeh, Christos Kaplanis, Shreya Pathak, Dharshan Kumaran, Anastasija Ilic, Jovana Mitrovic, Charles Blundell, and Andrea Banino. 2024. Synth 2 : Boosting visual-language models with synthetic captions and image embeddings. arXiv preprint arXiv:2403.07750

  23. [31]

    Ravi Shekhar, Sandro Pezzelle, Yauhen Klimovich, Aur \'e lie Herbelot, Moin Nabi, Enver Sangineto, and Raffaella Bernardi. 2017. Foil it! find one mismatch between image and language caption. arXiv preprint arXiv:1705.01359

  24. [32]

    Maria Tsimpoukelli, Jacob L Menick, Serkan Cabi, SM Eslami, Oriol Vinyals, and Felix Hill. 2021. Multimodal few-shot learning with frozen language models. Advances in Neural Information Processing Systems, 34:200--212

  25. [33]

    Thomas B Ward and Yuliya Kolomyts. 2010. Cognition and creativity. The Cambridge handbook of creativity, 5:93--112

  26. [34]

    Rui Xiao, Sanghwan Kim, Mariana-Iuliana Georgescu, Zeynep Akata, and Stephan Alaniz. 2024. Flair: Vlm with fine-grained language-informed image representations. arXiv preprint arXiv:2412.03561

  27. [35]

    Kaicheng Yang, Jiankang Deng, Xiang An, Jiawei Li, Ziyong Feng, Jia Guo, Jing Yang, and Tongliang Liu. 2023. Alip: Adaptive language-image pre-training with synthetic caption. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2922--2931

  28. [36]

    Kecheng Zheng, Yifei Zhang, Wei Wu, Fan Lu, Shuailei Ma, Xin Jin, Wei Chen, and Yujun Shen. 2024. Dreamlip: Language-image pre-training with long captions. In European Conference on Computer Vision, pages 73--90. Springer

  29. [37]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  30. [38]

    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 gl...

Pith tools

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