Pith. sign in

REVIEW 5 major objections 5 minor 49 references

PLPP: Prompt Learning with Perplexity Is Self-Distillation for Vision-Language Models

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A perplexity regularizer turns prompt learning into self-distillation and improves vision-language few-shot accuracy.

desk verdict PLPP is a clever, low-cost regularizer with a genuinely new penalty, but the reported gains are too small and too poorly quantified to establish superiority over PromptSRC. read the letter →

arxiv 2412.15277 v1 pith:YSLLEGOF submitted 2024-12-18 cs.CL cs.AI

Biao Liu , Wenyi Fang , Xiaoyu Wu , Yang Zheng , Zheng Hu , Bo Yuan This is my paper · ORCID
classification cs.CLcs.AI
keywords promptlearningperplexityself-distillationvision-languagemodelsCLIPfew-shotclassificationdomaingeneralizationregularization
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 proposes PLPP, a plug-in regularizer for prompt learning in vision-language models such as CLIP. It adds a perplexity loss that constrains the learnable prompt vectors toward the output distribution of a frozen text encoder, computed by matching prompt vectors to word embeddings for labels and running an untrained LM head for word probabilities. The authors show this perplexity objective is mathematically equivalent to hard-label self-distillation, and they soften the labels and truncate to top-k values to stabilize training and reduce computation. Across few-shot classification, base-to-novel generalization, cross-dataset evaluation, and domain generalization, PLPP improves over existing prompt-learning baselines without adding trainable parameters.

What carries the argument

The central object is the perplexity loss $L_{PPL}$ built from two distributions: $Q$, obtained by assigning each prompt vector the index of its nearest word embedding via cosine similarity (softened to a top-k distribution), and $P$, the word probability distribution produced by an untrained LM head, a linear layer without bias whose weights are initialized as the transpose of the embedding weights. The loss uses mutual self-distillation, $e^{\frac{1}{2}KL(Q_1\|P_1)} + e^{\frac{1}{2}KL(P_1\|Q_1)}$, plus an inverted term $L_{IPPL}$ computed from top-k selections in the opposite direction, and is added to the CLIP cross-entropy loss as $\lambda(\alpha L_{PPL} + (1-\alpha)L_{IPPL})$. This machinery ties prompt vectors to the vocabulary geometry of the frozen text encoder and delivers no-extra-parameter regularization.

What would settle it

Train PLPP on a dataset whose class names are absent from the model's vocabulary, such as invented or non-English tokens, and compare against PromptSRC; if the accuracy gains persist, the regularizer is not relying on vocabulary meaning, and if they vanish, vocabulary coverage is the load-bearing factor. A second check is to measure the KL divergence between Q and P under randomly initialized prompts; if it is already at chance level, the signal is degenerate.

Watch

Extended reading notes

Core claim

PLPP is a plug-in prompt-regularization method that makes the perplexity of a learnable prompt part of the training signal for frozen CLIP-style models. It obtains a label distribution Q by taking cosine similarities between prompt vectors and the embedding layer's word vectors, and a token distribution P by passing the encoded prompt through a linear LM head initialized from the transpose of the embedding weights; the regularizer is a symmetrized, exponentiated KL term between Q and P. The paper's conceptual claim is that this perplexity loss is equivalent to self-distillation, because it aligns the input prompt distribution with the text encoder's output distribution, and that replacing hard labels with top-k soft labels makes training stable and efficient. Experiments on four evaluation settings show consistent gains, with the clearest improvements over the PromptSRC baseline on novel-class generalization and domain-shift datasets.

Load-bearing premise

The regularizer stands on the premise that the untrained LM head, initialized from the transpose of the embedding weights, outputs a word-probability distribution whose top-k entries meaningfully match the prompt's intended meaning; if that head is not calibrated, the loss is not really perplexity and the reported gains could come from something else.

Editorial extensions

If this is right

  • Any existing prompt-learning method can be augmented with this perplexity regularizer without increasing the number of trainable parameters.
  • The top-k strategy keeps added computation small enough for practical few-shot and larger-scale use.
  • Improved base-to-novel and domain generalization imply that perplexity regularization reduces prompt overfitting to base-class patterns.
  • Because the loss is a form of self-distillation, the same idea can be applied at intermediate transformer layers rather than only at the final prompt.
  • In the reported experiments, PLPP improves over PromptSRC on most datasets for few-shot and novel-class accuracy, making it a drop-in upgrade for that baseline.

Reading between the lines

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

  • Because the LM head is tied to the embedding matrix, PLPP likely biases prompts toward the manifold of real word embeddings; one testable consequence is that learned prompts become more interpretable as token combinations than prompts trained without it.
  • The mutual KL can be read as a bidirectional distributional-consistency regularizer; a natural extension is to anneal the top-k size during training to balance stability and signal strength.
  • The method is defined for a fixed vocabulary, so applying it to multilingual or open-vocabulary CLIP variants would test whether vocabulary coverage of class names is what makes the regularizer effective.
  • If the self-distillation equivalence is right, the same perplexity objective should help prompt tuning in other frozen-encoder settings, including text-only language models, which the paper does not test.
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

5 major / 5 minor

Summary. The manuscript proposes PLPP, a plug-in regularization term for prompt learning in vision-language models, applied on top of PromptSRC. A perplexity-style loss is computed by (i) matching each learnable prompt vector to the nearest word in the frozen embedding layer via cosine similarity, and (ii) feeding the text through the frozen text encoder and an untrained linear LM head initialized from the transpose of the embedding weight, to obtain a word distribution. The regularizer is added to the standard cross-entropy prompt loss as two exponential KL terms, intended as perplexity and inverted perplexity losses. Experiments are reported on few-shot classification, base-to-novel generalization, cross-dataset transfer, and domain generalization on 11 datasets. The reported average gains over PromptSRC are small (e.g., +0.36 harmonic mean in base-to-novel, +0.1 average in domain generalization), and in cross-dataset evaluation PLPP's average (66.29) is below MaPLe (66.30).

Significance. The contribution is potentially useful if the empirical claims hold: the regularizer adds no trainable parameters, is conceptually simple, and the self-distillation interpretation is appealing. The manuscript is candid about settings and baselines and uses standard benchmarks. However, the main evidence is not statistically grounded, the loss as written is not the perplexity defined in the theory, and the top-k procedure is incompletely specified. As presented, the paper does not establish that PLPP reliably improves over PromptSRC; it provides suggestive but noisy evidence. The significance would be materially strengthened by variance reporting, significance tests, a corrected loss definition, and a complete specification of the top-k and hyperparameter choices.

major comments (5)
  1. [Prompt Learning with Perplexity, Eqs. (4)-(5)] The loss defined in Eqs. (4)-(5) is not the perplexity of Eq. (2). Equation (3) states that log-perplexity equals KL(Q||P), but Eq. (4) uses the exponential of one-half the KL divergence and symmetrizes the KL terms. The exponential of a KL divergence is not the perplexity, and the 1/2 factor and symmetrization are not motivated. If the final loss is only a heuristic regularizer, the paper should say so; if it is meant to implement perplexity, the definition must be corrected.
  2. [Prompt Learning with Perplexity, top-k description] The top-k truncation is underspecified. The text says the largest k values are retained in Q and the saved indexes are used to obtain P, but it does not state whether the truncated Q and P are renormalized to sum to one. KL divergence requires normalized distributions, so the behavior of the regularizer depends on this choice. Please specify the exact operation and, if renormalization is applied, state it explicitly.
  3. [Experiments, Tables 1-3 and Figure 2] The empirical support for the central claim is not statistically grounded. All numbers are averages over three seeds, but no standard deviations, confidence intervals, or paired significance tests are reported. The improvements over PromptSRC are small (harmonic mean +0.36 in Table 1, average +0.1 in Table 3, per-dataset few-shot gains of 0.1-1.3 in Figure 2), so they may be within seed noise. To support the claim of superior performance, report per-seed values or error bars and a paired test over the 11 datasets.
  4. [Implementation Details] The manuscript states that the best λ and α for few-shot and base-to-novel experiments are listed in the appendix, but no appendix is included. Per-dataset hyperparameter selection on test performance would make the comparisons optimistic and uncontrolled. Disclose the selected values and the selection protocol, or use fixed hyperparameters across datasets.
  5. [Prompt Learning with Perplexity, LM head initialization] The regularizer is only 'perplexity' if the untrained LM head produces a calibrated word distribution P. No calibration check is provided, and because Q and P are both derived from the same frozen text encoder, the self-distillation interpretation is true by construction rather than independent evidence of regularization value. A simple diagnostic, such as comparing the top-k probabilities with empirical token frequencies, would help establish that the loss has the claimed meaning.
minor comments (5)
  1. [Abstract and Conclusion] The abstract claims 'superior performance compared to existing methods,' but Table 2 shows PLPP's cross-dataset average (66.29) is below MaPLe's (66.30); the text acknowledges this, but the abstract and conclusion should be qualified accordingly.
  2. [Base-to-Novel Generalization] The phrase 'consistent performance advantages' is too strong because Table 1 shows PLPP's harmonic mean on FGVCAircraft (39.94) is below PromptSRC's (40.15), and its base accuracy is lower on four datasets.
  3. [Domain Generalization] The statement that PLPP 'consistently outperforms all competing methods across all target datasets' is inaccurate for ImageNet-R, where PLPP ties PromptSRC at 77.8, and for the source ImageNet accuracy, where PLPP (71.0) is below PromptSRC (71.3).
  4. [Throughout] There are frequent typographical errors, including 'PropmtSRC' in the Experiments section, 'retrain' for 'retain' in the top-k description, 'Moveover' in the Introduction, and 'data..' in Related Works.
  5. [Equations (3)-(5)] Equations (4) and (5) contain unmatched parentheses, e.g., 'e 1/2 ·KL(Q1||P1))' has an extra closing parenthesis; also 'H(Q, P) = logP P L' in Eq. (3) should be 'H(Q, P) = log PPL'.

Circularity Check

1 steps flagged · score 2.0 of 10

One self-definitional 'self-distillation' insight; empirical performance claims rest on external baselines and are not circular.

  1. self definitional [Methodology: Perplexity and Prompt Learning with Perplexity (Eq. 3-5)]
    "The Q distribution is calculated by using cosine similarity between prompts and embedding layer. As for P , we introduce an LM head positioned after the text encoder to output the distribution P . “The LM head consists of a simple linear layer without bias, with its weights initialized from the transpose of the embedding.weight."

    Q is derived from cosine similarity between the prompt vectors and the frozen embedding rows, while P is the output of the same frozen text encoder projected through a linear head whose weights are forced to be the transpose of those same embedding rows. Consequently, the KL terms in Eq. 4-5 compare two quantities that are both deterministic functions of one frozen embedding/encoder. The paper's 'unveiling' that PLPP is self-distillation is therefore a restatement of the construction (a model teaching itself) rather than an independently derived result. This is true by construction, but it is not load-bearing for the main empirical claim, which is evaluated against external baselines.

full rationale

The paper's central empirical claim is that adding PLPP improves few-shot, base-to-novel, cross-dataset, and domain-generalization performance over existing prompt-learning methods. That claim is supported by comparisons against external baselines (CoOp, CoCoOp, MaPLe, PromptSRC), which are not fitted by PLPP and are not self-citations. The only element that reduces to its own construction is the conceptual claim that perplexity in PLPP is self-distillation: Q and P are both produced from the same frozen embedding/text-encoder, so the mutual-KL regularizer enforces self-consistency by design. This is a true but tautological observation. The paper does not fit a parameter and then rename it as a prediction, and it does not rely on a self-citation chain to justify its method. The reported gains are small and lack error bars or significance tests, but statistical robustness is a correctness concern, not circularity. Overall, no significant circularity in the derivation chain; the empirical core is externally benchmarked.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central regularizer depends on several hand-chosen quantities (lambda, alpha, k, M) and on the untested validity of the no-training LM head and top-k truncation. The loss in Eq. 4-5 is not the perplexity of Eq. 2 because of the 1/2 exponent and the truncation, so the method rests on an under-specified and partly ad hoc construction.

free parameters (5)
  • loss weight lambda = 10 for cross-dataset and domain generalization; per-dataset best for few-shot and base-to-novel (appendix omitted)
    Eq. 6 weights the PLPP regularizer; results depend on per-dataset tuning in the missing appendix.
  • loss weight alpha = 0.2 for cross-dataset and domain generalization; per-dataset best elsewhere
    Eq. 6 balances LPPL and LIPPL; tuned per dataset without a reported validation protocol.
  • top-k k = 5
    Chosen for all tasks; truncates Q and P distributions before KL, affecting the loss value and gradient.
  • number of learnable prompt vectors M = 4
    Follows PromptSRC; the number of context vectors is a design choice that affects capacity.
  • exponent factor in LPPL and LIPPL = 1/2
    Eq. 4-5 define LPPL as e^{0.5 KL}, while Eq. 2 defines perplexity as e^H; the 1/2 factor is unexplained.
assumptions (4)
  • standard math For one-hot Q, H(Q)=0, so KL(Q||P)=H(Q,P)=log PPL.
    Eq. 3 relies on this identity; it is true for a one-hot distribution.
  • domain assumption Cosine similarity between each prompt vector and embedding rows yields meaningful word labels.
    The label distribution Q is defined this way in 'Prompt Learning with Perplexity'; no linguistic or semantic justification is given.
  • ad hoc to paper A linear layer initialized with the transpose of embedding.weight, without training, outputs a valid word probability distribution P.
    Used to define P in Eq. 4-5; the paper gives no normalization or calibration evidence.
  • ad hoc to paper Top-k truncation preserves enough distribution mass for KL(Q||P) to be a useful regularizer.
    Eq. 4-5 restrict Q and P to k=5 indices; no analysis of mass retained or effect on training is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PLPP: Prompt Learning with Perplexity Is Self-Distillation for Vision-Language Models." pith.science (2026). https://pith.science/paper/YSLLEGOF

@misc{pith2026241215277,
  author       = {Pith},
  title        = {Pith review of: PLPP: Prompt Learning with Perplexity Is Self-Distillation for Vision-Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YSLLEGOF}},
  note         = {Machine review of arXiv:2412.15277}
}
abstract

Pre-trained Vision-Language (VL) models such as CLIP have demonstrated their excellent performance across numerous downstream tasks. A recent method, Context Optimization (CoOp), further improves the performance of VL models on downstream tasks by introducing prompt learning. CoOp optimizes a set of learnable vectors, aka prompt, and freezes the whole CLIP model. However, relying solely on CLIP loss to fine-tune prompts can lead to models that are prone to overfitting on downstream task. To address this issue, we propose a plug-in prompt-regularization method called PLPP (Prompt Learning with PerPlexity), which use perplexity loss to regularize prompt learning. PLPP designs a two-step operation to compute the perplexity for prompts: (a) calculating cosine similarity between the weight of the embedding layer and prompts to get labels, (b) introducing a language model (LM) head that requires no training behind text encoder to output word probability distribution. Meanwhile, we unveil that the essence of PLPP is inherently a form of self-distillation. To further prevent overfitting as well as to reduce the additional computation introduced by PLPP, we turn the hard label to soft label and choose top-$k$ values for calculating the perplexity loss. For accelerating model convergence, we introduce mutual self-distillation learning, that is perplexity and inverted perplexity loss. The experiments conducted on four classification tasks indicate that PLPP exhibits superior performance compared to existing methods.

Figures

Figures reproduced from arXiv: 2412.15277 by the authors.

Figure 1
Figure 1. Overview of our proposed plug-in PLPP (Prompt Learning with PerPlexity) method for prompt learning in VL models. a self-regularizing framework, which includes mutual agree￾ment maximization, prompt self-ensembling regularization, and textual diversity regularization. Methodology In this section, we provide an overview of CoOp, and in￾troduce the evaluation metric perplexity in NLP and what its essence is in the prom… view at source ↗
Figure 2
Figure 2. The few-shot classification results on 11 datasets. We compare our PLPP with Liner probe CLIP, CoOp, MaPLe, and [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 41 canonical work pages

  1. [1]

    Bossard, L.; Guillaumin, M.; and Van Gool, L. 2014. Food-101--mining discriminative components with random forests. In ECCV, 446--461

    work page 2014

  2. [2]

    Chen, G.; Yao, W.; Song, X.; Li, X.; Rao, Y.; and Zhang, K. 2023. Prompt Learning with Optimal Transport for Vision-Language Models. In ICLR, 1--13

    work page 2023

  3. [3]

    Cimpoi, M.; Maji, S.; Kokkinos, I.; Mohamed, S.; and Vedaldi, A. 2014. Describing textures in the wild. In CVPR, 3606--3613

    2014

  4. [4]

    Deng, J.; Dong, W.; Socher, R.; Li, L.-J.; Li, K.; and Fei-Fei, L. 2009. Imagenet: A large-scale hierarchical image database. In CVPR, 248--255

    2009

  5. [5]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2021. An image is worth 16x16 words: Transformers for image recognition at scale. In ICLR, 1--13

    work page 2021

  6. [6]

    Fei-Fei, L.; Fergus, R.; and Perona, P. 2007. Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories. In CVIU, 59--70

    work page 2007

  7. [7]

    Gao, P.; Geng, S.; Zhang, R.; Ma, T.; Fang, R.; Zhang, Y.; Li, H.; and Qiao, Y. 2024. Clip-adapter: Better vision-language models with feature adapters. In IJCV, 581--595

    work page 2024

  8. [8]

    Gao, T.; Fisch, A.; and Chen, D. 2021. Making Pre-trained Language Models Better Few-shot Learners. In ACL, 3816--3830

    work page 2021

Show all 49 references
  1. [9]

    Geng, S.; Yuan, J.; Tian, Y.; Chen, Y.; and Zhang, Y. 2023. HiCLIP: Contrastive language-image pretraining with hierarchy-aware attention. In ICLR, 1--13

    2023

  2. [10]

    Guo, Z.; Zhang, R.; Qiu, L.; Ma, X.; Miao, X.; He, X.; and Cui, B. 2023. Calip: Zero-shot enhancement of clip with parameter-free attention. In AAAI, 746--754

    2023

  3. [11]

    X., Rui Zhang

    Hantao Yao, C. X., Rui Zhang. 2023. Visual-Language Prompt Tuning with Knowledge-guided Context Optimization. In CVPR, 6757--6767

    2023

  4. [12]

    Helber, P.; Bischke, B.; Dengel, A.; and Borth, D. 2018. Eurosat: A novel dataset and deep learning benchmark for land use and land cover classification. In IGARSS, 204--207

    2018

  5. [13]

    Hendrycks, D.; Basart, S.; Mu, N.; Kadavath, S.; Wang, F.; Dorundo, E.; Desai, R.; Zhu, T.; Parajuli, S.; Guo, M.; et al. 2021 a . The many faces of robustness: A critical analysis of out-of-distribution generalization. In ICCV, 8320--8329

    2021

  6. [14]

    Hendrycks, D.; Zhao, K.; Basart, S.; Steinhardt, J.; and Song, D. 2021 b . Natural adversarial examples. In CVPR, 15262--15271

    2021

  7. [15]

    Jang, J.; Kong, C.; Jeon, D.; Kim, S.; and Kwak, N. 2023. Unifying Vision-Language Representation Space with Single-Tower Transformer. In AAAI, 980--988

    2023

  8. [16]

    Jia, C.; Yang, Y.; Xia, Y.; Chen, Y.-T.; Parekh, Z.; Pham, H.; Le, Q.; Sung, Y.-H.; Li, Z.; and Duerig, T. 2021. Scaling up visual and vision-language representation learning with noisy text supervision. In ICML, 4904--4916

    2021

  9. [17]

    F.; Araki, J.; and Neubig, G

    Jiang, Z.; Xu, F. F.; Araki, J.; and Neubig, G. 2020. How can we know what language models know? In TACL, 423--438

    2020

  10. [18]

    Jin, W.; Cheng, Y.; Shen, Y.; Chen, W.; and Ren, X. 2022. A good prompt is worth millions of parameters: Low-resource prompt-based learning for vision-language models. In ACL, 2763--2775

    2022

  11. [19]

    U.; Rasheed, H.; Maaz, M.; Khan, S.; and Khan, F

    khattak, M. U.; Rasheed, H.; Maaz, M.; Khan, S.; and Khan, F. S. 2023. MaPLe: Multi-modal Prompt Learning. In CVPR, 19113--19122

    2023

  12. [20]

    U.; Wasim, S

    Khattak, M. U.; Wasim, S. T.; Naseer, M.; Khan, S.; Yang, M.-H.; and Khan, F. S. 2023. Self-regulating prompts: Foundational model adaptation without forgetting. In ICCV, 15190--15200

    2023

  13. [21]

    Krause, J.; Stark, M.; Deng, J.; and Fei-Fei, L. 2013. 3d object representations for fine-grained categorization. In ICCVW, 554--561

    2013

  14. [22]

    Lester, B.; Al-Rfou, R.; and Constant, N. 2021. The Power of Scale for Parameter-Efficient Prompt Tuning. In EMNLP, 3045--3059

    2021

  15. [23]

    L.; and Liang, P

    Li, X. L.; and Liang, P. 2021. Prefix-tuning: Optimizing continuous prompts for generation. In ACL, 4582--4597

    2021

  16. [24]

    Li, Y.; Liang, F.; Zhao, L.; Cui, Y.; Ouyang, W.; Shao, J.; Yu, F.; and Yan, J. 2022. Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm. In ICLR, 1--13

    2022

  17. [25]

    Lu, Y.; Liu, J.; Zhang, Y.; Liu, Y.; and Tian, X. 2022. Prompt distribution learning. In CVPR, 5206--5215

    2022

  18. [26]

    Maji, S.; Rahtu, E.; Kannala, J.; Blaschko, M.; and Vedaldi, A. 2013. Fine-grained visual classification of aircraft. arXiv preprint arXiv:1306.5151

    2013 arXiv

  19. [27]

    Nilsback, M.-E.; and Zisserman, A. 2008. Automated flower classification over a large number of classes. In ICVGIP, 722--729

    2008

  20. [28]

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

    Parkhi, O. M.; Vedaldi, A.; Zisserman, A.; and Jawahar, C. 2012. Cats and dogs. In CVPR, 3498--3505

    2012

  21. [29]

    Pillutla, K.; Swayamdipta, S.; Zellers, R.; Thickstun, J.; Welleck, S.; Choi, Y.; and Harchaoui, Z. 2021. MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontiers. In NeurIPS, 4816--4828

    2021

  22. [30]

    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 ICML, 8748--8763

    2021

  23. [31]

    Recht, B.; Roelofs, R.; Schmidt, L.; and Shankar, V. 2019. Do imagenet classifiers generalize to imagenet? In ICML, 5389--5400

    2019

  24. [32]

    L.; Wallace, E.; and Singh, S

    Shin, T.; Razeghi, Y.; Logan IV, R. L.; Wallace, E.; and Singh, S. 2020. Autoprompt: Eliciting knowledge from language models with automatically generated prompts. In EMNLP, 4222--4235

    2020

  25. [33]

    R.; and Shah, M

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

    2012 arXiv

  26. [34]

    Sun, J.; Qin, J.; Lin, Z.; and Chen, C. 2023. Prompt Tuning based Adapter for Vision-Language Model Adaption. arXiv preprint arXiv:2303.15234

    2023 arXiv

  27. [35]

    N.; and Kaiser

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; and Kaiser. 2017. Attention Is All You Need. In NeurIPS, 5998--6008

    2017

  28. [36]

    Wang, H.; Ge, S.; Lipton, Z.; and Xing, E. P. 2019. Learning robust global representations by penalizing local predictive power. In NeurIPS, 10506--10518

    2019

  29. [37]

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

    Xiao, J.; Hays, J.; Ehinger, K. A.; Oliva, A.; and Torralba, A. 2010. SUN database: Large-scale scene recognition from abbey to zoo. In CVPR, 3485--3492

    2010

  30. [38]

    Xing, Y.; Wu, Q.; Cheng, D.; Zhang, S.; Liang, G.; Wang, P.; and Zhang., Y. 2023. Dual Modality Prompt Tuning for Vision-Language Pre-Trained Model. In TMM, 1--13

    2023

  31. [39]

    Yang, C.; Meng, F.; Chen, S.; Liu, M.; and Zhang, R. 2023. Instance-Wise Adaptive Tuning and Caching for Vision-Language Models. In ECAI, 2834--2841

    2023

  32. [40]

    Yang, J.; Duan, J.; Tran, S.; Xu, Y.; Chanda, S.; Chen, L.; Zeng, B.; Chilimbi, T.; and Huang, J. 2022. Vision-Language Pre-Training with Triple Contrastive Learning. In CVPR, 15671--15680

    2022

  33. [41]

    You, H.; Zhou, L.; Xiao, B.; Codella, N.; Cheng, Y.; Xu, R.; Chang, S.-F.; and Yuan, L. 2022. Learning visual representation from modality-shared contrastive language-image pre-training. In ECCV, 69--87

    2022

  34. [42]

    Zhang, L.; Song, J.; Gao, A.; Chen, J.; Bao, C.; and Ma, K. 2019. Be your own teacher: Improve the performance of convolutional neural networks via self distillation. In ICCV, 3713--3722

    2019

  35. [43]

    Zhang, R.; Zhang, W.; Fang, R.; Gao, P.; Li, K.; Dai, J.; Qiao, Y.; and Li, H. 2022. Tip-adapter: Training-free adaption of clip for few-shot classification. In ECCV, 493--510

    2022

  36. [44]

    Zhong, Z.; Friedman, D.; and Chen, D. 2021. Factual probing is [mask]: Learning vs. learning to recall. In NAACL-HLT, 5017--5033

    2021

  37. [45]

    C.; and Liu, Z

    Zhou, K.; Yang, J.; Loy, C. C.; and Liu, Z. 2022 a . Conditional Prompt Learning for Vision-Language Models. In CVPR, 16795--16804

    2022

  38. [46]

    C.; and Liu, Z

    Zhou, K.; Yang, J.; Loy, C. C.; and Liu, Z. 2022 b . Learning to Prompt for Vision-Language Models. In IJCV, 2337--2348

    2022

  39. [47]

    Zhu, B.; Niu, Y.; Han, Y.; Wu, Y.; and Zhang, H. 2023. Prompt-aligned Gradient for Prompt Tuning. In ICCV, 15613--15623

    2023

  40. [48]

    , " * 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...

  41. [49]

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