Pith. sign in

REVIEW 3 major objections 6 minor 34 references

Classifier-Guided Captioning Across Modalities

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A frozen captioning model can be steered at inference time by a text classifier, and the paper shows this raises zero-shot audio captioning to state-of-the-art results on AudioCaps and Clotho.

desk verdict A clean inference-time guidance idea with a circular audibility metric and a text-only classifier that doesn't actually test whether captions are audible. read the letter →

arxiv 2501.03183 v1 pith:U6X664MR submitted 2025-01-03 cs.CL cs.AIcs.SDeess.AS

Ariel Shaulov , Tal Shaharabany , Eitan Shaar , Gal Chechik , Lior Wolf This is my paper · ORCID
classification cs.CLcs.AIcs.SDeess.AS
keywords audiocaptioningzero-shotclassifierguidanceinference-timeoptimizationaudibilityGPT-4syntheticdatacontextcachemodalityadaptation
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 claims that a captioning model can be adapted to a new modality's semantics without any retraining, by steering its frozen language model at inference time with a text classifier. The target application is zero-shot audio captioning, where the desired semantics is audibility—describing only events that can actually be heard, rather than visually inferred actions like walking. The authors train a binary audibility classifier on 10,000 sentences generated by ChatGPT and use its gradient to adjust the transformer's context cache during decoding, with a cross-entropy term that keeps the output close to the original language model. Combined with the existing NoAudioCaptioning system, this guidance raises every reported captioning metric on AudioCaps and Clotho in both transfer directions, which the authors describe as state-of-the-art zero-shot audio captioning. The same recipe also improves two image captioning models, suggesting the mechanism is modality-agnostic.

What carries the argument

The mechanism is gradient-guided context cache optimization. During autoregressive decoding, the model keeps its weights fixed and instead adjusts the key and value vectors in the transformer's context cache, starting from the frozen model's own cache. The update minimizes a weighted sum of a cross-entropy loss, which anchors the next-token distribution to the unmodified language model, and a classifier loss, which pushes the whole sentence toward the audibility label. The audibility classifier is a DistilBERT trained on GPT-4-generated examples of audible and non-audible sentences.

What would settle it

A human rating study would settle the central claim: give listeners captions generated by NoAudioCaptioning with and without the guidance, from the same audio clips, and ask which are more audible. If human raters choose the unguided captions at chance level or prefer them, or if the guidance-induced gains on the Aud metric do not reproduce with a classifier trained on human-labeled captions, the central claim would be refuted.

Watch

Extended reading notes

Core claim

The central claim is that inference-time classifier guidance can inject modality-specific semantics into a frozen captioning system. For audio, the authors define audibility as the property that a caption describes only sounds directly present in the clip, and show that optimizing the context cache of a GPT-2-based zero-shot captioner against an audibility classifier's score systematically improves caption quality. They report that NoAudioCaptioning plus their guidance outperforms all compared zero-shot baselines and also beats the out-of-domain supervised baselines on most metrics, in both transfer directions, AudioCaps to Clotho and Clotho to AudioCaps. They interpret this as evidence that a single training-free guidance loop can substitute for task-specific fine-tuning.

Load-bearing premise

Everything rests on the assumption that a classifier trained on ChatGPT-generated examples of 'audible' versus 'not audible' sentences is a reliable judge of whether a real audio caption describes sounds a listener can actually hear.

Editorial extensions

If this is right

  • Guidance improves NoAudioCaptioning on every reported metric in both transfer directions; on AudioCaps to Clotho, Audibility Accuracy rises from 59.8 to 78.2, and on Clotho to AudioCaps from 65.6 to 85.7.
  • With guidance, the zero-shot system reaches or exceeds the out-of-domain performance of fully supervised audio captioners on most metrics, despite never seeing paired audio-text data.
  • The same inference-time loop lifts the image captioners ZeroCap and ClipCap on MS-COCO, including the reference-free CLIP-S metric, indicating the classifier can be swapped to steer different semantic properties.
  • Because no training of the captioner is involved, the approach can be deployed with a frozen model and swapped classifiers for different objectives.

Reading between the lines

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

  • If the same optimization were applied with a classifier for another semantic property, such as politeness, technicality, or factual consistency, it could steer a frozen captioner toward that property without architectural changes.
  • The reported Aud metric is computed with the same classifier that provides the guidance, so a human listening study would be needed to confirm that the guidance makes captions genuinely more audible rather than more similar to synthetic training text.
  • The roughly two-second-per-token cost suggests the method is best suited to offline captioning; reducing the number of optimization steps or distilling the guidance into a lightweight prefix could make it practical for interactive use.
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. This paper proposes an inference-time classifier-guidance framework for captioning. A frozen captioning model (NoAudioCaptioning with GPT-2) generates text autoregressively; at each step the context cache is optimized by gradient descent to minimize a weighted sum of a cross-entropy regularization term and a classifier loss that maximizes an audibility score for the generated sentence. The audibility classifier is a DistilBERT binary classifier trained on 10k GPT-4-generated sentences labeled as audible or non-audible. The authors evaluate on AudioCaps and Clotho in an out-of-domain zero-shot setting, reporting consistent improvements over NoAudioCaptioning on standard captioning metrics, CLAP-S, BERT-S, and a newly introduced Audibility Accuracy (Aud) metric, and they claim state-of-the-art zero-shot audio captioning. An additional image-captioning experiment with ZeroCap and ClipCap on MS-COCO is presented as an ablation showing generality.

Significance. The framework is attractive because it requires no retraining of the captioning model, is conceptually simple, and the reported standard-metric improvements are consistent across two transfer directions and across modalities. If the audibility gains reflect true perceptual properties, the method would be a useful plug-in for adapting captioners to new semantic requirements. However, the central audibility claim is currently supported primarily by a metric computed with the same classifier that is optimized during generation, and the classifier has no audio input; the paper therefore needs external validation before the contribution can be accepted.

major comments (3)
  1. [Sec. IV, Table II and Sec. III-A, Eq. (1)] The Aud metric used in Table II is computed with the same audibility classifier h_a that is optimized during inference in Eq. (1). The large improvements on Aud (59.8 to 78.2 on AudioCaps to Clotho; 65.6 to 85.7 on Clotho to AudioCaps) are therefore expected consequences of optimizing the classifier's own score, not independent evidence that captions are more audible. Please report an audibility evaluation based on human listening or on a classifier or automatic metric that was not used as the optimization target.
  2. [Sec. III-A and Eq. (1)] The audibility classifier is trained solely on 10k GPT-4-generated text sentences, with no audio input and no pairing with audio recordings. As described, such a classifier can at best learn a text prior about which words describe potentially audible events; it has no mechanism to verify that a described event is actually present in a given audio clip. Because Eq. (1) optimizes this text-only score, the reported Aud gains and the claim that the method adapts captions to the audio modality are not established. The paper should include a listening study or an audio-grounded evaluation, such as a classifier trained on real audio-caption pairs or a retrieval-based metric against the clip, to demonstrate that the guidance improves true per-clip audibility.
  3. [Table II] No error bars, confidence intervals, or significance tests are reported, although several standard-metric gains are small (e.g., BLEU4 from 7.0 to 7.7 on AudioCaps to Clotho; METEOR from 12.7 to 13.5 on Clotho to AudioCaps). Without repeated runs or statistical testing, the claim that the method consistently improves captioning quality is not fully supported. Please provide variance across at least three seeds or a significance test for the main comparisons.
minor comments (6)
  1. [Eq. (1)] The expression h_a(LM(x_i; C_i)[1]) is ambiguous; please clarify whether the classifier operates on the full generated sentence or on the next-token distribution, and what [1] indexes.
  2. [Sec. IV, Implementation Details] The text says the optimization process employs batch size 64, learning rate 0.0003, and 40 epochs, but the inference-time optimization is per-token and cannot run for 40 epochs; please disambiguate classifier training from inference-time optimization.
  3. [Sec. IV, Implementation Details] The target sequence length is fixed to 30 tokens for all AudioCaps and Clotho captions, while reference captions vary in length; please justify this choice or report sensitivity to the sequence length.
  4. [Table I] The example 'Ringing phone awaits an answer' reads as a narrative caption rather than a canonical audio description; replacing it with a clearer audible example would make the dataset description easier to interpret.
  5. [Throughout] The terms 'ChatGPT' and 'GPT-4' are used interchangeably; please use consistent terminology when describing the dataset generation.
  6. [References] The related work section calls [24] 'NoAudioCaptioning', but the reference list entry is 'Training audio captioning models without audio'; please ensure the citation naming is consistent.

Circularity Check

1 steps flagged · score 6.0 of 10

The Audibility Accuracy metric is computed with the same classifier that Eq. 1 optimizes, so the reported Aud gains are self-referential; the other standard metrics remain independent evidence.

  1. self definitional [Section IV, Implementation details; Eq. (1); Section III-A; Table II]
    "We evaluate our pipeline using standard captioning metrics BLEU [13], METEOR [14], ROUGE-L [15], SPICE [16], CIDEr [17], semantic metrics CLAP-S [34] for audio-caption alignment and BERT-S [33] for semantic similarity with BERT embeddings, and introduce Audibility Accuracy (Aud) using our pre-trained audibility classifier (see Sec. III-A)."

    The optimization objective in Eq. (1) is L_classifier = -log(h_a(LM(x_i; C_i)[1])), which directly increases the audibility probability assigned by the same classifier h_a that is later used to compute the Aud metric in Table II. The Aud jumps (59.8 to 78.2 and 65.6 to 85.7) are therefore largely a consequence of gradient ascent on the evaluation function itself, not evidence from an independent audibility label, human annotation, or audio-grounded signal. Since h_a is trained only on GPT-4-generated text (Sec. III-A) and never conditions on audio, 'audibility' is defined by the synthetic text distribution, and the same definition is used as both the guidance target and the measurement.

full rationale

The paper's central generation mechanism is an inference-time optimization of a frozen captioning model's context cache using two losses: a classifier loss L_classifier (Eq. 1) and a cross-entropy regularizer (Eq. 2). The captioning model itself is not retrained, and the reported improvements on standard reference-based and semantic metrics are measured against external benchmark references and CLAP/BERT embeddings, so those results are not circular. The circular component is specifically the audibility evaluation: the paper introduces Audibility Accuracy (Aud) computed 'using our pre-trained audibility classifier,' and Eq. 1 optimizes the log-probability of that same classifier. Thus the Aud metric is not an independent test of audibility; it is the objective function under a different name. This makes the audibility-specific claim partially self-definitional, but not the whole paper. I found no load-bearing self-citation chain or imported uniqueness argument; the cited baselines are external prior work. The missing audio grounding and lack of human listening evaluation are validity concerns, not circularity per se, and I did not count them as circular steps.

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

The central claim depends on the audibility classifier as an oracle and on several hand-set hyperparameters; no new physical entities are introduced.

free parameters (3)
  • Loss weights lambda0 and lambda1 = 0.2 and 0.6
    Set by hand as default parameters in Eq. 3; no sensitivity analysis is reported.
  • Inference optimization learning rate and number of steps n
    Not reported in the paper; these control how strongly the context cache is moved toward the classifier and are needed to reproduce results.
  • Target sequence length = 30 tokens
    Chosen for evaluation; affects generation and reported metrics.
assumptions (4)
  • domain assumption The audibility classifier's output is a valid measure of whether a caption describes audible content.
    The entire guidance and the Aud metric rely on this classifier, which is trained on GPT-4 generated text rather than human-audited real audio captions (Sec. III-A, Eq. 1).
  • domain assumption GPT-4 generated sentences provide a representative distribution of audible and non-audible captions for real audio events.
    The classifier is trained only on synthetic text; real audio captions may have different distributions (Sec. III-A).
  • domain assumption Gradient updates to the context cache, regularized by cross-entropy, preserve language model fluency.
    The method assumes the editable context cache (Eq. 1 and Eq. 2) can be locally adjusted without breaking LM coherence.
  • domain assumption Standard captioning metrics (BLEU, METEOR, CIDEr, CLAP-S, BERT-S) are adequate proxies for caption quality.
    Tables II and III use these metrics without human evaluation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Classifier-Guided Captioning Across Modalities." pith.science (2026). https://pith.science/paper/U6X664MR

@misc{pith2026250103183,
  author       = {Pith},
  title        = {Pith review of: Classifier-Guided Captioning Across Modalities},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U6X664MR}},
  note         = {Machine review of arXiv:2501.03183}
}
read the original abstract

Most current captioning systems use language models trained on data from specific settings, such as image-based captioning via Amazon Mechanical Turk, limiting their ability to generalize to other modality distributions and contexts. This limitation hinders performance in tasks like audio or video captioning, where different semantic cues are needed. Addressing this challenge is crucial for creating more adaptable and versatile captioning frameworks applicable across diverse real-world contexts. In this work, we introduce a method to adapt captioning networks to the semantics of alternative settings, such as capturing audibility in audio captioning, where it is crucial to describe sounds and their sources. Our framework consists of two main components: (i) a frozen captioning system incorporating a language model (LM), and (ii) a text classifier that guides the captioning system. The classifier is trained on a dataset automatically generated by GPT-4, using tailored prompts specifically designed to enhance key aspects of the generated captions. Importantly, the framework operates solely during inference, eliminating the need for further training of the underlying captioning model. We evaluate the framework on various models and modalities, with a focus on audio captioning, and report promising results. Notably, when combined with an existing zero-shot audio captioning system, our framework improves its quality and sets state-of-the-art performance in zero-shot audio captioning.

Figures

Figures reproduced from arXiv: 2501.03183 by the authors.

Figure 1
Figure 1. An overview of our modality-agnostic approach is depicted, with a focus on audio captioning during inference. In this illustration, the audio modality is [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 25 canonical work pages

  1. [1]

    Show and tell: A neural image caption generator,

    O. Vinyals, A. Toshev, S. Bengio, and D. Erhan, “Show and tell: A neural image caption generator,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 3156–3164

    2015

  2. [2]

    Deep visual-semantic alignments for generating image descriptions,

    A. Karpathy and L. Fei-Fei, “Deep visual-semantic alignments for generating image descriptions,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 3128–3137

    2015

  3. [3]

    Long-term recurrent convolutional networks for vi- sual recognition and description,

    J. Donahue et al., “Long-term recurrent convolutional networks for vi- sual recognition and description,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2015, pp. 2625–2634

    work page 2015

  4. [4]

    Show, attend and tell: Neural image caption generation with visual attention,

    K. Xu et al. , “Show, attend and tell: Neural image caption generation with visual attention,” in International conference on machine learning , 2015, pp. 2048–2057

    work page 2015

  5. [5]

    Image captioning with semantic attention,

    Q. You, H. Jin, Z. Wang, C. Fang, and J. Luo, “Image captioning with semantic attention,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 4651–4659

    2016

  6. [6]

    Bottom-up and top-down attention for image captioning and visual question answering,

    P. Anderson et al. , “Bottom-up and top-down attention for image captioning and visual question answering,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2018, pp. 6077–6086

    work page 2018

  7. [7]

    Automated audio caption- ing: Describing audio content with natural language,

    K. Drossos, T. Lipping, and T. Virtanen, “Automated audio caption- ing: Describing audio content with natural language,” in 2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), 2017, pp. 374–378

    work page 2017

  8. [8]

    Audio captioning using pre-trained large-scale language model guided by audio-based similarity,

    Y . Koizumi et al. , “Audio captioning using pre-trained large-scale language model guided by audio-based similarity,” in ICASSP 2020- 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 814–818

    work page 2020

Show all 34 references
  1. [9]

    Audiocaps: Generating captions for audios in the wild,

    H. Kim et al., “Audiocaps: Generating captions for audios in the wild,” in Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics (NAACL) , 2019

    2019

  2. [10]

    Audio caption: Listen and tell,

    Y . Wu, Y . Liu, and Y . Liu, “Audio caption: Listen and tell,” in Proceedings of the 27th ACM International Conference on Multimedia , 2019, pp. 2629–2637

    2019

  3. [11]

    DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter,

    V . Sanh, L. Debut, J. Chaumond, and T. Wolf, “DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter,” in arXiv preprint arXiv:1910.01108, 2019

    1910 arXiv

  4. [12]

    Clotho: An audio captioning dataset,

    K. Drossos, S. Lipping, and T. Virtanen, “Clotho: An audio captioning dataset,” in ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2020, pp. 7364– 7368

    2020

  5. [13]

    BLEU: A method for automatic evaluation of machine translation,

    K. Papineni, S. Roukos, T. Ward, and W. Zhu, “BLEU: A method for automatic evaluation of machine translation,” in Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics , 2002, pp. 311–318

    2002

  6. [14]

    METEOR: An automatic metric for MT evaluation with improved correlation with human judgments,

    S. Banerjee and A. Lavie, “METEOR: An automatic metric for MT evaluation with improved correlation with human judgments,” in Pro- ceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization, 2005, pp. 65– 72

    2005

  7. [15]

    ROUGE: A package for automatic evaluation of summaries,

    C. Y . Lin, “ROUGE: A package for automatic evaluation of summaries,” in Text Summarization Branches Out: Proceedings of the ACL-04 Workshop, 2004, pp. 74–81

    2004

  8. [16]

    SPICE: Semantic Propositional Image Caption Evaluation,

    C. Y . Lin, “SPICE: Semantic Propositional Image Caption Evaluation,” in Computer Vision–ECCV 2016: 14th European Conference, Proceed- ings, Part V 14 , 2016, pp. 382–398

    2016

  9. [17]

    CIDEr: Consensus-based image description evaluation,

    R. Vedantam, C. L. Zitnick, and D. Parikh, “CIDEr: Consensus-based image description evaluation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 4566–4575, 2015

    2015

  10. [18]

    CLIPScore: A reference-free evaluation metric for image captioning,

    J. Hessel, A. Holtzman, M. Forbes, R. L. Bras, and Y . Choi, “CLIPScore: A reference-free evaluation metric for image captioning,” in arXiv preprint arXiv:2104.08718, 2021

    2021 arXiv

  11. [19]

    ZeroCap: Zero-shot image-to-text generation for visual-semantic arithmetic,

    Y . Tewel, S. Shalev, O. Pineau, G. Chechik, and O. Levy, “ZeroCap: Zero-shot image-to-text generation for visual-semantic arithmetic,” in Advances in Neural Information Processing Systems (NeurIPS) , 2022

    2022

  12. [20]

    CLIPCap: CLIP prefix for image captioning,

    R. Mokady, A. Hertz, and A. Hoffer, “CLIPCap: CLIP prefix for image captioning,” in arXiv preprint arXiv:2111.09734 , 2021

    2021 arXiv

  13. [21]

    Prefix tuning for automated audio captioning,

    M. Kim, K. Sung-Bin, and T.-H. Oh, “Prefix tuning for automated audio captioning,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2023

    2023

  14. [22]

    RECAP: retrieval-augmented audio captioning,

    S. Ghosh, S. Kumar, C. K. R. Evuru, R. Duraiswami, and D. Manocha, “RECAP: retrieval-augmented audio captioning,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Pro- cessing (ICASSP), 2024, pp. 1161–1165

    2024

  15. [23]

    Zero-shot audio captioning with audio-language model guidance and audio context keywords,

    L. Salewski, S. Fauth, A. Koepke, and Z. Akata, “Zero-shot audio captioning with audio-language model guidance and audio context keywords,” arXiv preprint arXiv:2311.08396 , 2023

    2023 arXiv

  16. [24]

    Training audio captioning models without audio,

    S. Deshmukh, B. Elizalde, D. Emmanouilidou, B. Raj, R. Singh, and H. Wang, “Training audio captioning models without audio,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 371–375

    2024

  17. [25]

    Weakly-supervised automated audio captioning via text only training,

    T. Kouzelis and V . Katsouros, “Weakly-supervised automated audio captioning via text only training,” arXiv preprint arXiv:2309.12242 , 2023

    2023 arXiv

  18. [26]

    Zero-Shot Audio Captioning Using Soft and Hard Prompts,

    Y . Zhang, X. Xu, R. Du, H. Liu, Y . Dong, Z.-H. Tan, W. Wang, and Z. Ma, “Zero-Shot Audio Captioning Using Soft and Hard Prompts,” in arXiv preprint arXiv:2406.06295 , 2024

    2024 arXiv

  19. [27]

    Show and Tell: A Neural Image Caption Generator,

    Oriol Vinyals and Alexander Toshev and Samy Bengio and Dumitru Erhan, “Show and Tell: A Neural Image Caption Generator,”In Proceed- ings of the IEEE conference on computer vision and pattern recognition 2015

    2015

  20. [28]

    Show, Attend and Tell: Neural Image Caption Gen- eration with Visual Attention,

    Kelvin Xu and Jimmy Ba and Ryan Kiros and Kyunghyun Cho and Aaron Courville and Ruslan Salakhutdinov and Richard Zemel and Yoshua Bengio, “Show, Attend and Tell: Neural Image Caption Gen- eration with Visual Attention,” In International conference on machine learning, 2015

    2015

  21. [29]

    Self-critical Sequence Training for Image Captioning,

    Steven J. Rennie, Etienne Marcheret, Youssef Mroueh, Jarret Ross and Vaibhava Goel, “Self-critical Sequence Training for Image Captioning,” In Proceedings of the IEEE conference on computer vision and pattern recognition, 2017

    2017

  22. [30]

    Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering,

    Peter Anderson and Xiaodong He and Chris Buehler and Damien Teney and Mark Johnson and Stephen Gould and Lei Zhang, “Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering,” In Proceedings of the IEEE conference on computer vision and pattern recog...

    2018

  23. [31]

    Microsoft COCO: Common objects in context,

    T. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll´ar, and C. L. Zitnick, “Microsoft COCO: Common objects in context,” in European Conference on Computer Vision (ECCV) , pp. 740–755, 2014

    2014

  24. [32]

    Flickr30k Entities: Collecting region-to-phrase corre- spondences for richer image-to-sentence models,

    B. Plummer, L. Wang, C. M. Cervantes, J. C. Caicedo, J. Hockenmaier, and S. Lazebnik, “Flickr30k Entities: Collecting region-to-phrase corre- spondences for richer image-to-sentence models,” in Proceedings of the IEEE International Conference on Computer Vision (ICCV) , pp. 26...

    2015

  25. [33]

    BERTScore: Evaluating text generation with BERT,

    T. Zhang, V . Kishore, F. Wu, K. Q. Weinberger, and Y . Artzi, “BERTScore: Evaluating text generation with BERT,” arXiv preprint arXiv:1904.09675, 2019

    1904 arXiv

  26. [34]

    CLAP: Learning audio concepts from natural language supervision,

    B. Elizalde, S. Deshmukh, M. Al Ismail, and H. Wang, “CLAP: Learning audio concepts from natural language supervision,” in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 1–5, 2023

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.