Pith. sign in

REVIEW 6 major objections 5 minor 41 references

CM3T: Framework for Efficient Multimodal Learning for Inhomogeneous Interaction Datasets

T0 review · 6 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read CM3T claims frozen video transformers can be retuned by plug-in adapters using only 12.8% of backbone parameters for video — 22.3% with two extra modalities — matching or beating full finetuning on three interaction datasets.

desk verdict Useful PETL framework with honest empirical claims, but the spatial-chunking explanation for the multi-head adapter is not isolated by the experiments. read the letter →

arxiv 2501.03332 v1 pith:T2PLZ6CN submitted 2025-01-06 cs.CV

Tanay Agrawal , Mohammed Guermal , Michal Balazia , Francois Bremond This is my paper · ORCID
classification cs.CV
keywords parameter-efficienttransferlearningmulti-headvisionadaptercross-attentionprefixtuningvideoclassificationmultimodalSwinTransformerinteractionrecognition
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 claims that adapter-based transfer learning, which normally only succeeds when the video backbone was pretrained with expensive self-supervised methods such as CLIP or VideoMAE, can be made to work with ordinary supervised checkpoints instead. Its framework, CM3T, freezes a Video Swin Transformer pretrained on Kinetics-400 and attaches two new kinds of trainable plugins: multi-head vision adapters, which give different groups of spatial windows their own adapter heads, and cross-attention adapters, which fold audio, optical flow, or text into the model as additional plugins. With a frozen backbone, CM3T reports 48.2% top-1 accuracy on Epic-Kitchens-100 (full finetuning reaches 48.9%, the three-branch state of the art 49.6%), a state-of-the-art mean MSE of 0.69 on UDIVA, and 0.901 mAP on MPIIGroupInteraction, all while training 12.8% of the backbone's parameters for video alone and 22.3% for two extra modalities. A sympathetic reader would care because it points to a way of reusing the widely available, cheap-to-produce supervised checkpoints instead of forcing users toward self-supervised pretraining.

What carries the argument

The load-bearing object is the multi-head vision adapter: a scaled parallel adapter whose input is divided along the spatial window (patch) dimension, with separate down/up projections per head, so that each frame region gets its own fitted transformation, plus a learnable scaling factor $s$ that lets the adapter change activations more forcefully. It is what lets a frozen, supervised-pretrained backbone absorb a new dataset's distribution shift. The second mechanism is the cross-attention adapter, which replaces the adapter's linear layers with cross-attention over a side modality's embedding ($Q = V = \mathrm{ReLU}(xW_{\mathrm{down}})$, $K$ = modality embedding), turning a new modality into a self-contained trainable plugin; AdapterFusion then weights the per-modality outputs so the modalities are merged rather than summed. Prefix tuning, kept as the third module, is modified with ReLU activation and gated addition.

What would settle it

Permute which spatial windows are assigned to which adapter head and retrain: if the head-wise split still performs as well, the gain is just added parameters and the spatial premise is not doing the work. Alternatively, run the same MHVA-versus-plain-adapter comparison (39.8% versus 28.7% on EK-100) on a dataset with heavy camera or object motion, such as Something-Something, where the fixed-camera assumption breaks; matching performance would refute the claimed mechanism.

Watch

Extended reading notes

Core claim

The central claim is that the poor performance of adapters on supervised-pretrained video transformers is not a limit of the adapter idea but of its shape. Existing scaled parallel adapters put the same linear down/up projection on every spatial patch, whereas CM3T splits the input along the window dimension of the Video Swin transformer and gives each spatial chunk its own adapter head with a smaller bottleneck, so different frame regions learn different representations of the dataset shift; swapping in this design lifts the same prefix-tuned recipe from 28.7% to 39.8% on Epic-Kitchens-100. The paper further claims that adding a modality is a pure plug-in operation: a cross-attention adapter whose queries and values come from the vision bottlenecks attends to the new modality's embedding as key, and AdapterFusion merges the per-modality outputs. With three input sources, CM3T gains 8.4 points over its RGB-only version on EK-100, whereas the much larger M&M Mix ensemble gains only 2.9 points over its own RGB-only base. On a second backbone, ViViT-B, the same plugins surpass full finetuning (44.3% versus 37.4%), which the paper reads as evidence that the method is model-agnostic.

Load-bearing premise

The multi-head vision adapter's advantage rests on interaction videos having roughly fixed cameras so that the same spatial regions reliably hold the same content; without that regularity, or if the pretraining data is too dissimilar to the target, the method has no demonstrated edge over plain adapters.

Editorial extensions

If this is right

  • Supervised checkpoints, not only self-supervised ones, become usable for parameter-efficient transfer: a frozen Video Swin-B pretrained on Kinetics-400 comes within 0.7 points of full finetuning on EK-100 while training 12.8% of the parameters.
  • Each new modality is a standalone plug-in costing about 5.8M extra trainable parameters, and the added modalities are put to use efficiently: CM3T gains 8.4 points on EK-100 when two modalities join, more than the 2.9-point gain of the much larger M&M Mix ensemble.
  • Adapter weights transfer across datasets for the same modality, which the paper reports speeds convergence on small datasets: MPIIGI converges in 9 epochs where full finetuning needs 17.
  • The recipe transfers across architectures: the same plugins on ViViT-B reach 44.3% top-1, surpassing that backbone's full finetuning at 37.4%, supporting the framework's model-agnostic claim.

Reading between the lines

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

  • A control the paper does not run would separate mechanism from parameter count: randomly shuffle which spatial windows each adapter head processes; if accuracy barely changes, the claimed spatial prior is not what carries the gain.
  • If the fixed-camera premise is the real driver, the head-wise split should transfer to other geometry-stable domains such as fixed-surveillance video or medical imaging, and its edge over plain adapters should shrink on mobile or freely moving-camera footage, a boundary that could be mapped directly.
  • Because the cross-attention adapter only needs an embedding of the side modality, the same 5.8M-parameter plug-in likely extends to other signals the paper never tries — depth, gaze, skeleton pose, or language instructions — making the framework a generic fusion socket rather than a video-specific trick.
  • The learnable scaling factor and ReLU-based prefix tuning suggest a training-dynamics story of easier convergence on small data; that predicts the largest advantage in the low-data regime, which could be checked by subsampling the training sets.
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

6 major / 5 minor

Summary. The paper presents CM3T, a parameter-efficient transfer learning framework for video transformers. It adds multi-head vision adapters (MHVA) and cross-attention adapters (CAA) to a frozen backbone, together with an approximation of prefix tuning. The authors evaluate on Epic-Kitchens-100, MPIIGroupInteraction, and UDIVA v0.5, reporting accuracies comparable to full fine-tuning while training only 12.8% (video) or 22.3%/28.9% (multimodal) of the backbone parameters.

Significance. If validated, the central claim would be practically valuable: CM3T would make supervised-pretrained video transformers adaptable with a small number of trainable parameters, without requiring self-supervised pretraining. The paper tests the approach on three datasets with different tasks and two backbones, and includes comparisons to full fine-tuning and to other PETL methods. However, the key architectural novelty (MHVA) is not isolated by a matched-parameter ablation, and the SOTA comparisons are partly confounded. The efficiency numbers are encouraging but the causal explanation of the gain is not yet supported.

major comments (6)
  1. [§3.2.2, §4.2.1] The central claim that the multi-head vision adapter (MHVA) is what enables strong performance with supervised pretraining is not established by the evidence. The only direct comparison is between 'MHV A + PT' (39.8%) and 'Adaptformer + PT' (28.7%) in Table 1, but these two configurations differ in three ways: head-splitting, learnable scaling factor, and the total number of adapters/parameters. The paper does not ablate the head-splitting alone: a single-head adapter with the same total bottleneck dimension, the same learnable scaling, and the same placement would isolate the effect of spatial chunking. Without this matched-parameter control, the improvement cannot be attributed to the multi-head design.
  2. [§1, §3.2.2] The stated mechanism for MHVA is that interaction datasets have almost fixed cameras and stable spatial locations of objects and body parts, so splitting the input along the window dimension allows each head to specialize. This explanation is inconsistent with the implementation: in Video Swin, window partitions shift between consecutive blocks, so a patch is not assigned to a fixed spatial head across layers. Moreover, EK-100 is egocentric and the camera moves, yet MHVA still yields a large gain. The paper should either provide a different rationale or directly test the spatial-specialization hypothesis (e.g., compare fixed versus shifted window assignment, or random head assignment). As written, the offered mechanism is not supported.
  3. [§4.2.1, Table 1] The SOTA comparison with M&M Mix is confounded by the use of a different backbone architecture and pretraining protocol. The claim of 'comparable and even better results than the state-of-the-art' (abstract) is only true with a CLIP backbone, whereas the 12.8% parameter-efficiency claim refers to Video Swin-B. Please separate these two claims and report a direct comparison where the backbone is held fixed (e.g., full fine-tuning of the same Video Swin-B).
  4. [§4.2.3, Table 4] The percentage of trainable parameters for multimodal input is 22.3% in EK-100 but 28.9% in UDIVA and MPIIGI. The abstract's statement that 'only 22.3% trainable parameters for two additional modalities' is therefore not accurate across the three datasets. Please report all configurations consistently and specify the reference backbone for each percentage.
  5. [§3.2, §4] The paper does not report several hyperparameters that are necessary to reproduce the experiments: bottleneck dimension r, number of heads in MHVA, initial scaling factor s, gated addition factor λ, dropout values, optimizer, learning rate, and number of training seeds. Without these, the proposed method cannot be evaluated or reproduced from the text. At minimum, the values used in the main experiments should be listed in a table.
  6. [§4.2.1, Table 1] No variance or number of seeds is reported for the main results. On small datasets such as MPIIGI, the differences between methods (e.g., 0.901 vs 0.899 mAP) are likely within the noise of a single run. The paper should report mean and standard deviation over multiple runs, or at least clarify that the numbers are single runs.
minor comments (5)
  1. [Throughout] There are several typos, including 'UDIV A v0.5' (should be UDIVA) and 'tanH' (should be tanh).
  2. [§3.3, Eqs. (7)-(10)] The notation for AdapterFusion is unclear: s' is first defined as a softmax over a row, then used as a scalar multiplier; please clarify dimensions and the concatenation in Eq. (9).
  3. [Figure 2] Figure 2 is dense and the three adapter blocks are hard to distinguish; consider enlarging or adding a schematic of the MHVA head-splitting.
  4. [§3.3] The sentence 'the cross-attention module proposed by [2] is used in place of the traditional one' is not reflected in the equations. Please specify how the module from [2] is incorporated and how it reduces parameters.
  5. [§3.2.1, Eqs. (1)-(5)] Several symbols in the equations are not defined (e.g., Wq, Wk, Wv, C, λ). Please define all symbols in the text or in a table.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: central claims are tested on external benchmarks; only a minor non-load-bearing self-citation appears among SOTA baselines.

full rationale

CM3T's central claims are not derived from its own definitions. The main evidence for the multi-head vision adapter is an empirical comparison on EK-100 between scaled parallel adapters with prefix tuning (28.7%) and MHVA with prefix tuning (39.8%, Table 1), plus comparisons to full finetuning (Table 4), M&M Mix (Table 1), ST-adapters and Dual-path adapters (Table 1), and ViViT experiments (Table 6). The equations in Section 3.2 define the adapter blocks without assuming the target result. The learnable scaling factor is a fitted parameter, but the paper reports a fixed-scaling ablation (s=4.0 gives 1.1% lower on EK-100) rather than presenting the learned value as a prediction. The only self-citation that appears in the evaluation is the authors' own FAt transformers [2] as the SOTA baseline for UDIVA and MPIIGI (Section 4.2.2, Tables 2-3) and MMCA [2] as a cross-attention alternative (Table 5). That is a baseline-selection weakness, not a circular reduction: the comparison does not define CM3T in terms of [2]'s outputs, and the same claims are also supported by non-self baselines such as M&M Mix, full finetuning, AdaptFormer, and ST-adapter. No uniqueness theorem or ansatz is imported through self-citation, and no known result is merely renamed. The skeptic's concern that the fixed-camera explanation is not supported by a matched ablation is a correctness/evidence question, not circularity. Overall, the derivation chain is self-contained; the score reflects the minor, non-load-bearing self-citation in the state-of-the-art comparison.

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

The framework introduces no new physical or conceptual entities; it reuses adapters, prefix tuning, cross-attention, and AdapterFusion. The main unstated inputs are architectural hyperparameters rather than free parameters fitted to data, though the initial scaling factor s is tuned by the authors (4.0 vs. learnable) and is a hand-chosen value that affects the EK-100 result.

free parameters (4)
  • bottleneck dimension r = not reported
    The adapter bottleneck size is a hand-chosen hyperparameter that sets the parameter count and capacity in Equations (1)-(2); the paper only says a smaller dimension was used.
  • number of heads in MHVA = not reported
    The multi-head vision adapter splits the input along the window dimension into {h1, h2, ...}; the number of splits is a design choice not disclosed.
  • initial scaling factor s = 4.0 (one comparison); otherwise learned
    Section 3.2.2 says a learnable scaling factor improves results; fixing it at 4.0 loses 1.1% on EK-100, so its initialization is relevant but not reported.
  • gated addition factor λ = not reported
    Equation (3) uses λ for prefix tuning gated addition; the paper does not state whether λ is learned or fixed, or its value.
assumptions (3)
  • domain assumption A frozen supervised-pretrained backbone provides features relevant to the downstream task.
    The entire PETL approach assumes the pretrained model already captures the basic spatial and motion understanding needed for the target videos (Section 3.2).
  • domain assumption Interaction datasets have a stable spatial layout across the video.
    The multi-head vision adapter's benefit is motivated by fixed cameras and body parts appearing at consistent locations (Section 1); this is asserted, not measured.
  • standard math Standard matrix algebra and softmax attention behave as defined.
    Equations (1)-(10) assume standard attention, softmax, and linear algebra operations; no formal verification is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CM3T: Framework for Efficient Multimodal Learning for Inhomogeneous Interaction Datasets." pith.science (2026). https://pith.science/paper/T2PLZ6CN

@misc{pith2026250103332,
  author       = {Pith},
  title        = {Pith review of: CM3T: Framework for Efficient Multimodal Learning for Inhomogeneous Interaction Datasets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T2PLZ6CN}},
  note         = {Machine review of arXiv:2501.03332}
}
read the original abstract

Challenges in cross-learning involve inhomogeneous or even inadequate amount of training data and lack of resources for retraining large pretrained models. Inspired by transfer learning techniques in NLP, adapters and prefix tuning, this paper presents a new model-agnostic plugin architecture for cross-learning, called CM3T, that adapts transformer-based models to new or missing information. We introduce two adapter blocks: multi-head vision adapters for transfer learning and cross-attention adapters for multimodal learning. Training becomes substantially efficient as the backbone and other plugins do not need to be finetuned along with these additions. Comparative and ablation studies on three datasets Epic-Kitchens-100, MPIIGroupInteraction and UDIVA v0.5 show efficacy of this framework on different recording settings and tasks. With only 12.8% trainable parameters compared to the backbone to process video input and only 22.3% trainable parameters for two additional modalities, we achieve comparable and even better results than the state-of-the-art. CM3T has no specific requirements for training or pretraining and is a step towards bridging the gap between a general model and specific practical applications of video classification.

Figures

Figures reproduced from arXiv: 2501.03332 by the authors.

Figure 1
Figure 1. This is a representation of the main problem CM3T aims [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Detailed architecture of CM3T. Colored parts are the ones that are finetuned and the rest are frozen. It has three separate blocks [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 27 canonical work pages

  1. [1]

    Multimodal personality recognition using cross-attention transformer and behaviour encoding

    Tanay Agrawal, Dhruv Agarwal, Michal Balazia, Neelabh Sinha, and Francois Bremond. Multimodal personality recognition using cross-attention transformer and behaviour encoding. In Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 5: VISAPP , pages 501–508. INSTICC, SciT...

    work page 2022

  2. [2]

    Multimodal vision transformers with forced attention for behavior analysis

    Tanay Agrawal, Michal Balazia, Philipp M ¨uller, and Franc ¸ois Br´emond. Multimodal vision transformers with forced attention for behavior analysis. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 3392–3402, January 2023. 6, 7, 8

    work page 2023

  3. [3]

    V ATT: transformers for multimodal self-supervised learning from raw video, audio and text

    Hassan Akbari, Liangzhe Yuan, Rui Qian, Wei-Hong Chuang, Shih-Fu Chang, Yin Cui, and Boqing Gong. V ATT: transformers for multimodal self-supervised learning from raw video, audio and text. CoRR, abs/2104.11178, 2021. 3

    arXiv 2021

  4. [4]

    Vivit: A video vi- sion transformer

    Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lu ˇci´c, and Cordelia Schmid. Vivit: A video vi- sion transformer. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision (ICCV), pages 6836– 6846, October 2021. 2

    work page 2021

  5. [5]

    Bodily be- haviors in social interaction: Novel annotations and state-of- the-art evaluation

    Michal Balazia, Philipp M ¨uller, ´Akos Levente T´anczos, Au- gust von Liechtenstein, and Franc ¸ois Br´emond. Bodily be- haviors in social interaction: Novel annotations and state-of- the-art evaluation. In Proceedings of the 30th ACM Inter- national Conference on Multimedia, pages 70–79, 2022. 6, 7

    work page 2022

  6. [6]

    Adaptformer: Adapt- ing vision transformers for scalable visual recognition.arXiv preprint arXiv:2205.13535, 2022

    Shoufa Chen, Chongjian Ge, Zhan Tong, Jiangliu Wang, Yibing Song, Jue Wang, and Ping Luo. Adaptformer: Adapt- ing vision transformers for scalable visual recognition.arXiv preprint arXiv:2205.13535, 2022. 3, 5, 6, 7

    arXiv 2022

  7. [7]

    Rescaling egocentric vision

    Dima Damen, Hazel Doughty, Giovanni Maria Farinella, Antonino Furnari, Evangelos Kazakos, Jian Ma, Davide Moltisanti, Jonathan Munro, Toby Perrett, Will Price, et al. Rescaling egocentric vision. arXiv preprint arXiv:2006.13256, 2020. 6

    arXiv 2006

  8. [8]

    A transformer-based joint-encoding for emotion recognition and sentiment analysis

    Jean-Benoit Delbrouck, No ´e Tits, Mathilde Brousmiche, and St´ephane Dupont. A transformer-based joint-encoding for emotion recognition and sentiment analysis. In Second Grand-Challenge and Workshop on Multimodal Language (Challenge-HML), pages 1–7, Seattle, USA, July 2020. As- sociation for Computational Linguistics. 3

    work page 2020

Show all 41 references
  1. [9]

    Ppt: Pre-trained prompt tuning for few-shot learning

    Yuxian Gu, Xu Han, Zhiyuan Liu, and Minlie Huang. Ppt: Pre-trained prompt tuning for few-shot learning. arXiv preprint arXiv:2109.04332, 2021. 3

    2021 arXiv

  2. [10]

    Towards a unified view of parameter-efficient transfer learning

    Junxian He, Chunting Zhou, Xuezhe Ma, Taylor Berg- Kirkpatrick, and Graham Neubig. Towards a unified view of parameter-efficient transfer learning. arXiv preprint arXiv:2110.04366, 2021. 2, 3, 4, 5

    2021 arXiv

  3. [11]

    Parameter-efficient trans- fer learning for NLP

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. Parameter-efficient trans- fer learning for NLP. In Kamalika Chaudhuri and Ruslan Salakhutdinov, editors,Proceedings of the 36th Inte...

    2019

  4. [12]

    Lora: Low-rank adaptation of large language models, 2021

    Edward Hu, Yelong Shen, Phil Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models, 2021. 3

    2021

  5. [13]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021. 3

    2021 arXiv

  6. [14]

    Mumu: Cooperative mul- titask learning-based guided multimodal fusion

    Md Mofijul Islam and Tariq Iqbal. Mumu: Cooperative mul- titask learning-based guided multimodal fusion. In Proceed- ings of the AAAI Conference on Artificial Intelligence , vol- ume 36(1), pages 1043–1051, 2022. 3

    2022

  7. [15]

    Vi- sual prompt tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Vi- sual prompt tuning. In Computer Vision–ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXXIII, pages 709–727. Springer, 2022. 3

    2022

  8. [16]

    Compacter: Efficient low-rank hypercomplex adapter layers

    Rabeeh Karimi Mahabadi, James Henderson, and Sebastian Ruder. Compacter: Efficient low-rank hypercomplex adapter layers. Advances in Neural Information Processing Systems, 34:1022–1035, 2021. 3

    2021

  9. [17]

    Parameter-efficient multi-task fine-tuning for transformers via shared hypernetworks

    Rabeeh Karimi Mahabadi, Sebastian Ruder, Mostafa De- hghani, and James Henderson. Parameter-efficient multi-task fine-tuning for transformers via shared hypernetworks. In Annual Meeting of the Association for Computational Lin- guistics, 2021. 3

    2021

  10. [18]

    Transformers in vision: A survey

    Salman Khan, Muzammal Naseer, Munawar Hayat, Syed Waqas Zamir, Fahad Shahbaz Khan, and Mubarak Shah. Transformers in vision: A survey. ACM computing surveys (CSUR), 54(10s):1–41, 2022. 3

    2022

  11. [19]

    Gated mechanism for attention based multimodal sentiment analysis

    Ayush Kumar and Jithendra Vepa. Gated mechanism for attention based multimodal sentiment analysis. CoRR, abs/2003.01043, 2020. 3

    2003 arXiv

  12. [20]

    The power of scale for parameter-efficient prompt tuning

    Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning. In Proceed- ings of the 2021 Conference on Empirical Methods in Nat- ural Language Processing , pages 3045–3059, Online and Punta Cana, Dominican Republic, Nov. 2021. Assoc...

    2021

  13. [21]

    Prefix-tuning: Optimiz- ing continuous prompts for generation

    Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimiz- ing continuous prompts for generation. arXiv preprint arXiv:2101.00190, 2021. 3

    2021 arXiv

  14. [22]

    Video swin transformer

    Ze Liu, Jia Ning, Yue Cao, Yixuan Wei, Zheng Zhang, Stephen Lin, and Han Hu. Video swin transformer. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 3202–3211, June

  15. [23]

    Parameter-efficient multi-task fine-tuning for transformers via shared hypernetworks.arXiv preprint arXiv:2106.04489, 2021

    Rabeeh Karimi Mahabadi, Sebastian Ruder, Mostafa De- hghani, and James Henderson. Parameter-efficient multi-task fine-tuning for transformers via shared hypernetworks.arXiv preprint arXiv:2106.04489, 2021. 3

    2021 arXiv

  16. [24]

    Unipelt: A unified framework for parameter-efficient lan- guage model tuning

    Yuning Mao, Lambert Mathias, Rui Hou, Amjad Almahairi, Hao Ma, Jiawei Han, Wen-tau Yih, and Madian Khabsa. Unipelt: A unified framework for parameter-efficient lan- guage model tuning. arXiv preprint arXiv:2110.07577 ,

    arXiv

  17. [25]

    Tiny adapters for vision transformers, 2023

    Imad Eddine Marouf, Enzo Tartaglione, and St ´ephane Lath- uili`ere. Tiny adapters for vision transformers, 2023. 3, 5

    2023

  18. [26]

    Context- aware personality inference in dyadic scenarios: Introducing the udiva dataset

    Cristina Palmero, Javier Selva, Sorina Smeureanu, Julio Ju- nior, CS Jacques, Albert Clap ´es, Alexa Mosegu ´ı, Zejian Zhang, David Gallardo, Georgina Guilera, et al. Context- aware personality inference in dyadic scenarios: Introducing the udiva dataset. In Proceedings of the...

  19. [27]

    St-adapter: Parameter-efficient image-to-video transfer learning for action recognition

    Junting Pan, Ziyi Lin, Xiatian Zhu, Jing Shao, and Hong- sheng Li. St-adapter: Parameter-efficient image-to-video transfer learning for action recognition. arXiv preprint arXiv:2206.13559, 2022. 1, 3, 6, 7

    2022 arXiv

  20. [28]

    Dual-path adaptation from image to video transformers

    Jungin Park, Jiyoung Lee, and Kwanghoon Sohn. Dual-path adaptation from image to video transformers. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 2203–2213, June 2023. 1, 3, 6, 7

    2023

  21. [29]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In Proceedings of th...

    2021

  22. [30]

    Learning multiple visual domains with residual adapters.Ad- vances in neural information processing systems , 30, 2017

    Sylvestre-Alvise Rebuffi, Hakan Bilen, and Andrea Vedaldi. Learning multiple visual domains with residual adapters.Ad- vances in neural information processing systems , 30, 2017. 3

    2017

  23. [31]

    Efficient parametrization of multi-domain deep neural net- works

    Sylvestre-Alvise Rebuffi, Hakan Bilen, and Andrea Vedaldi. Efficient parametrization of multi-domain deep neural net- works. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018. 3

    2018

  24. [32]

    Multi- lingual detection of check-worthy claims using world lan- guages and adapter fusion.arXiv preprint arXiv:2301.05494,

    Ipek Baris Schlicht, Lucie Flek, and Paolo Rosso. Multi- lingual detection of check-worthy claims using world lan- guages and adapter fusion.arXiv preprint arXiv:2301.05494,

    arXiv

  25. [33]

    Aman Shenoy and Ashish Sardana. Multilogue-net: A context-aware rnn for multi-modal emotion detection and sentiment analysis in conversation.Second Grand-Challenge and Workshop on Multimodal Language (Challenge-HML) ,

  26. [34]

    Vl-adapter: Parameter-efficient transfer learning for vision-and-language tasks

    Yi-Lin Sung, Jaemin Cho, and Mohit Bansal. Vl-adapter: Parameter-efficient transfer learning for vision-and-language tasks. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 5227–5237,

  27. [35]

    Training neu- ral networks with fixed sparse masks

    Yi-Lin Sung, Varun Nair, and Colin A Raffel. Training neu- ral networks with fixed sparse masks. Advances in Neural Information Processing Systems, 34:24193–24205, 2021. 3

    2021

  28. [36]

    VideoMAE: Masked autoencoders are data-efficient learners for self-supervised video pre-training

    Zhan Tong, Yibing Song, Jue Wang, and Limin Wang. VideoMAE: Masked autoencoders are data-efficient learners for self-supervised video pre-training. InAdvances in Neural Information Processing Systems, 2022. 1

    2022

  29. [37]

    Videomae: Masked autoencoders are data-efficient learn- ers for self-supervised video pre-training

    Zhan Tong, Yibing Song, Jue Wang, and Limin Wang. Videomae: Masked autoencoders are data-efficient learn- ers for self-supervised video pre-training. arXiv preprint arXiv:2203.12602, 2022. 3, 5

    2022 arXiv

  30. [38]

    M&m mix: A multimodal multiview transformer ensemble

    Xuehan Xiong, Anurag Arnab, Arsha Nagrani, and Cordelia Schmid. M&m mix: A multimodal multiview transformer ensemble. arXiv preprint arXiv:2206.09852, 2022. 2, 3, 6, 7, 8

    2022 arXiv

  31. [39]

    Multiview transformers for video recognition

    Shen Yan, Xuehan Xiong, Anurag Arnab, Zhichao Lu, Mi Zhang, Chen Sun, and Cordelia Schmid. Multiview transformers for video recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3333–3343, 2022. 8

    2022

  32. [40]

    Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models

    Elad Ben Zaken, Shauli Ravfogel, and Yoav Goldberg. Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models. arXiv preprint arXiv:2106.10199,

  33. [41]

    Chengqing Zong, Fei Xia, Wenjie Li, and Roberto Navigli. Proceedings of the 59th annual meeting of the association for computational linguistics and the 11th international joint conference on natural language processing (volume 1: Long papers). In Proceedings of the 59th Annua...

    2021

Pith tools

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