Pith. sign in

REVIEW 3 major objections 5 minor 43 references

Separating Shared and Domain-Specific LoRAs for Multi-Domain Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper argues that forcing the shared LoRA into the column space of the pretrained weights and each domain-specific LoRA into the left null space guarantees the two kinds of adapters learn different information, with small and mixed…

desk verdict Clean subspace-orthogonality idea for multi-domain LoRA, but the 'left null space' guarantee is approximate and the empirical support is too thin for full acceptance. read the letter →

arxiv 2508.02978 v1 pith:RVIJSU4Z submitted 2025-08-05 cs.CV

Yusaku Takama , Ning Ding , Tatsuya Yokota , Toru Tamaki This is my paper
classification cs.CV
keywords multi-domainlearninglow-rankadaptationLoRAsubspaceseparationcolumnspaceleftnulltruncatedSVDactionrecognition
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 tries to establish that in multi-domain fine-tuning with LoRA, the shared and domain-specific adapters can be made structurally distinct rather than only trained on different data. The proposed method pins the shared LoRA updates to the column space of each frozen pretrained weight and pins each domain's LoRA updates to the corresponding left null space, obtained by truncating the singular value decomposition at a 95% cumulative contribution threshold. On action recognition with a frozen ImageNet-pretrained Vision Transformer, the constraints give small gains in some configurations (UCF101 92.69 to 92.88 with subspace constraints; HMDB51 66.05 to 66.45 with subspace constraints plus domain separation) while slightly lowering accuracy on other datasets. The authors present the result as a structural guarantee worth further investigation, not as a new accuracy record.

What carries the argument

The central object is the truncated singular value decomposition of each frozen pretrained linear weight $W = U\Sigma V^{\top}$, with the cutoff chosen so that the top singular vectors accumulate 95% of the squared singular values and form the column space $\mathrm{col}(W)$; the remaining singular vectors span the manufactured left null space $\mathrm{Ker}(W^{\top})$. Projection matrices $P_m = U_m U_m^{\top}$ and $P_n = U_n U_n^{\top}$ are inserted into the forward pass, so that a layer computes $h = Wx + P_m\Delta W x + P_n\Delta W_i x$. Because the gradient of each LoRA weight is multiplied by the same symmetric projection, every parameter update stays inside its assigned subspace. A second mechanism is a pair of losses on each domain-specific factor $B_i$: an orthonormality loss $\lVert B_i^{\top}B_i - I_r\rVert_F^2$ and a subspace-separation loss that maximizes the Frobenius distance between $B_iB_i^{\top}$ and $B_jB_j^{\top}$, pushing different domains onto different bases within the left null space.

What would settle it

Train the same upper-heavy multi-domain model with the projections swapped: domain-specific LoRAs confined to the strong column-space directions of $W$ and the shared LoRA confined to the left null space, then compare UCF101, Kinetics400, and HMDB51 accuracy against the paper's configuration. If the swapped assignment does not clearly degrade accuracy, the paper's claim that shared knowledge lives in $\mathrm{col}(W)$ and domain-specific knowledge in $\mathrm{Ker}(W^T)$ is unsupported.

Watch

Extended reading notes

Core claim

The central claim is that the ambiguity of shared-plus-domain-specific LoRA architectures—where a shared adapter trains on all domains and a domain-specific adapter trains on one domain, without any explicit difference in what they are allowed to represent—can be removed by construction. The paper claims that restricting the shared LoRA $\Delta W$ to the column space $\mathrm{col}(W)$ of the pretrained weight and each domain-specific LoRA $\Delta W_i$ to the left null space $\mathrm{Ker}(W^T)$ makes the two types of updates mutually orthogonal, so the shared module cannot silently absorb domain-specific information and the domain-specific module cannot overwrite pretrained knowledge. Because the pretrained weights are full rank, the left null space is manufactured by truncated SVD: singular directions below a 95% cumulative contribution are treated as zero and assigned to the domain-specific side. A further pair of losses pushes each domain's LoRA basis toward orthonormality and away from every other domain's basis. The empirical claim is that this enforced separation helps in some action-recognition configurations and hurts in others, and that the LoRA dimension analysis confirms the separation changes how the adapters use their rank.

Load-bearing premise

The argument stands on the assumption that the singular directions of the pretrained weight $W$ below the 95% cumulative contribution cutoff carry no useful pretrained knowledge, and that all domain-agnostic information lives in the strong column space; if either part fails, the projection discards useful adaptations instead of separating them.

Editorial extensions

If this is right

  • If the orthogonality guarantee holds, adding or updating a domain-specific LoRA cannot change the shared representation, which should reduce interference in multi-domain training.
  • The cumulative-contribution analysis of LoRA weights gives a diagnostic: with subspace constraints alone, domain-specific LoRAs concentrate in fewer effective dimensions, and adding the separation losses spreads them across all 32 rank dimensions.
  • The 95% truncation threshold becomes a tunable hyperparameter with measurable consequences; on HMDB51, threshold 0.80 outperforms 0.95 by more than one accuracy point.
  • The method scales spatially: with the left null space roughly 20–80% of each layer's dimension and about 600 dimensions available at rank 32, three domains are far from exhausting the space, so more domains could be separated without extra parameters beyond one LoRA per domain.
  • The accuracy gains reported are small and inconsistent, so the paper's contribution is best read as a structural separation guarantee plus an analysis method, not as a large performance improvement over prior multi-domain fine-tuning.

Reading between the lines

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

  • The same subspace-splitting recipe could be applied to other parameter-efficient adapters, such as residual adapters or prefix vectors, whenever a pretrained weight supplies a column space; the paper does not test these settings.
  • A direct check of the load-bearing assumption would be to swap the projection: confine domain-specific LoRAs to the top principal directions and the shared LoRA to the low-variance directions; if accuracy does not clearly fall, the paper's identification of which subspace carries shared knowledge is wrong.
  • The static-bias motivation suggests a testable extension: measure whether domain-specific LoRAs confined to the left null space reduce a model's reliance on static scene features in action recognition compared with unconstrained LoRAs.
  • As the number of domains grows, the fixed left null space dimension will eventually saturate; the paper's own dimension curves imply that the useful width of each domain-specific LoRA shrinks with more domains, so a scaling test with ten or more domains would clarify the approach's practical ceiling.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a method for multi-domain learning with LoRA adapters that explicitly separates shared and domain-specific information by constraining shared LoRA updates to the column space of each pretrained weight matrix W and domain-specific LoRA updates to a subspace treated as the left null space of W, using projection matrices derived from SVD. It further introduces two auxiliary losses (Lorth and Lss) to encourage orthonormality and separation among domain-specific LoRA bases. The method is evaluated on action recognition across UCF101, Kinetics400, and HMDB51 with a frozen ImageNet-21K-pretrained ViT, using two integration structures (upper-heavy and all-flat). Results show small and inconsistent accuracy differences compared to unconstrained variants, and an analysis of the effective dimensions of LoRA weights is presented.

Significance. The core idea is attractive: if shared and domain-specific LoRA updates could be confined to mutually orthogonal subspaces by construction, it would provide a principled way to prevent interference in multi-domain PEFT. The gradient projection derivation in Eqs. (5)-(7) is correct, and the projection matrices are derived from the frozen pretrained weights (not learned), which is a strength. However, the paper's central 'by construction' claim is not exact, because the pretrained weights are full-rank and the method actually uses a truncated SVD tail rather than a true left null space. Moreover, the experimental evidence is not yet sufficient to establish effectiveness: no multiple seeds or variance are reported, differences are within a few tenths of a point with mixed signs, and no existing multi-domain PEFT methods are included as baselines. The paper is honest about these limitations (Sections 3.1.1 and 4.2.3), but as it stands the main theoretical virtue is an unvalidated spectral assumption and the empirical support is fragile.

major comments (3)
  1. [Section 3.1.2 and Section 3.2] The central claim that domain-specific LoRAs are confined to Ker(W^T) and therefore 'do not affect the features obtained from W' is not exact as stated. For full-rank pretrained weight matrices, the true left null space is trivial (for square or overdetermined layers), so the method in Section 3.1.2 truncates the SVD at a 95% cumulative contribution threshold and renames the tail span U_n as the left null space. The discarded directions have nonzero singular values, so W U_n is not zero and a domain-specific LoRA in U_n does change the model output along directions the pretrained network actively uses. The claimed non-interference guarantee holds only to the extent that the discarded 5% of squared singular energy is semantically irrelevant, which is an empirical assumption, not a construction. The paper's own Table 3 shows that for HMDB51 the lowest threshold (0.80) gives the best performance, which is hard to reconcile with the rationale that a higher threshold better preserves pretrained knowledge. The authors should either prove a bound on the interference, empirically validate the spectral assumption, or re-frame the contribution as an approximate separation rather than an exact one.
  2. [Section 4.2, Tables 1-3] The empirical support for effectiveness is weak and inconsistent. In Table 1, with the upper-heavy structure, adding subspace constraints improves UCF101 by only +0.19 and leaves HMDB51 unchanged, while degrading Kinetics400 by -0.29; adding both constraints improves HMDB51 by +0.40 but degrades UCF101 by -0.95 and Kinetics400 by -0.33. In the all-flat structure, the proposed constraints mostly hurt performance. No multiple seeds, standard deviations, or significance tests are reported, so these differences may be within noise. Furthermore, there is no comparison to MTLoRA, VMT-Adapter, or any other existing multi-domain PEFT method; the paper only compares variants of its own method. Without stronger experiments, the claim that the proposed separation is effective in practice is not established.
  3. [Section 3.3, Eq. (9)] The separation loss Lss = - (1/sqrt(2)) sum_{i<j} ||B_i B_i^T - B_j B_j^T||_F is proposed to push domain-specific subspaces apart, but the paper provides no analysis of how this loss interacts with the orthonormality loss Lorth or with the scale of the B_i matrices. Since ||B_i B_i^T||_F is not normalized, the negative distance could be trivially affected by scaling, and the reported configuration (lambda1=1, lambda2=1e-7) is the only one tested. Table 1 does not isolate the effect of Lss from the subspace projection, so it is unclear whether the separation loss contributes to the observed results. A sensitivity study or a proof that the loss has the intended geometric effect is needed before the domain-separation claim can be accepted.
minor comments (5)
  1. [Section 3.2] The sentence 'Then we introduce constrains the LoRA weights...' contains a grammatical error; 'constrains' should be 'constraints'.
  2. [Section 3.3, Eq. (9)] The text after Eq. (9) refers to 'Bd1 and Bd2' but the formula uses 'Bi and Bj'; please unify the notation.
  3. [Section 4.2.4] The text says 'The curve colors show three domains (UCF100, Kinetics400, and HMDB51)' but the dataset is UCF101; please correct the typo.
  4. [References] Reference [10] has 'yelong shen' in lowercase; please fix the capitalization to 'Yelong Shen'.
  5. [Section 4.2.2] The text states 'In general, performance improves as r increases' but Table 2 shows HMDB51 peaking at r=16 and declining at r=32; please clarify this statement to account for the non-monotonic behavior.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the subspace constraints are a construction evaluated on held-out accuracy, not a fitted quantity masquerading as a prediction.

full rationale

The paper's claimed contribution is methodological: it constructs projection matrices P_m and P_n from the SVD of the frozen pre-trained weights and constrains LoRA gradient updates to orthogonal subspaces via Equations (5)-(7). The reported performance numbers are measured on held-out validation sets of UCF101, Kinetics400, and HMDB51, so no accuracy result is forced by construction. Hyperparameters such as rank r, the 0.95 cumulative-contribution threshold, and lambda values are selected using the same benchmarks, but this is ordinary model selection rather than definitional circularity: the validation accuracies are not algebraic consequences of the fitted constants. The conceptual weakness identified in Section 3.1.2 and Section 4.2.3, namely that the truncated SVD tail is called the left null space even though the pretrained weights are full-rank, is a spectral approximation and an unvalidated assumption about which singular directions carry task-relevant information; the paper explicitly acknowledges the threshold dependence ('This observation needs more investigation'). Importantly, the paper does not claim to discover that LoRAs naturally separate, nor does it derive the non-interference guarantee from data: it proposes the separation as a constraint and evaluates it empirically. No load-bearing claim is justified by a self-citation; MTLoRA and TesNet are cited as baselines or inspirations, not as the paper's own prior results. Therefore no specific circular reduction can be exhibited, and the appropriate finding is no significant circularity.

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

The method's central claim rests on an assumed decomposition of pretrained knowledge into a top singular subspace and a leftover low-variance subspace, plus several manually chosen hyperparameters. No new physical or model entities are introduced.

free parameters (4)
  • cumulative contribution threshold (C_k) = 0.95
    Chosen by hand; determines the split between column space and left null space for every linear layer. Table 3 shows it changes HMDB accuracy by more than one point.
  • LoRA rank r = 32
    Set after sweeping r=4,8,16,32,64 in Table 2; UCF and Kinetics favor 32, while HMDB favors 16.
  • orthogonality loss weight lambda_1 = 1
    Fixed hyperparameter in the final loss Eq. (10); no ablation is reported.
  • domain-separation loss weight lambda_2 = 1e-7
    Fixed hyperparameter in Eq. (10); no ablation is reported.
assumptions (3)
  • domain assumption The pretrained weight W is full rank but can be approximated by truncating singular values below a 95% cumulative contribution threshold, creating a zero subspace.
    Used in Section 3.1.2 to construct col(W) and Ker(W^T); because W is full rank, the left null space is only null for the truncated matrix.
  • ad hoc to paper The column space of W stores shared, domain-agnostic knowledge; the left null space is safe for domain-specific information and does not interfere with pretrained knowledge.
    Central motivation in Sections 1 and 3.2; not proven and threshold dependent.
  • domain assumption An ImageNet-pretrained ViT provides a transferable body for action recognition, and separating video-specific LoRAs can address static bias.
    Used in Sections 1 and 4.1; no experiment directly measures static bias reduction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Separating Shared and Domain-Specific LoRAs for Multi-Domain Learning." pith.science (2026). https://pith.science/paper/RVIJSU4Z

@misc{pith2026250802978,
  author       = {Pith},
  title        = {Pith review of: Separating Shared and Domain-Specific LoRAs for Multi-Domain Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RVIJSU4Z}},
  note         = {Machine review of arXiv:2508.02978}
}
read the original abstract

Existing architectures of multi-domain learning have two types of adapters: shared LoRA for all domains and domain-specific LoRA for each particular domain. However, it remains unclear whether this structure effectively captures domain-specific information. In this paper, we propose a method that ensures that shared and domain-specific LoRAs exist in different subspaces; specifically, the column and left null subspaces of the pre-trained weights. We apply the proposed method to action recognition with three datasets (UCF101, Kinetics400, and HMDB51) and demonstrate its effectiveness in some cases along with the analysis of the dimensions of LoRA weights.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

43 extracted references · 35 canonical work pages

  1. [1]

    MT- LoRA: Low-rank adaptation approach for efficient multi- task learning

    Ahmed Agiza, Marina Neseem, and Sherief Reda. MT- LoRA: Low-rank adaptation approach for efficient multi- task learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16196–16205, 2024. 1, 2, 3, 5

    work page 2024

  2. [2]

    Multi-task learning with deep neural networks: A survey

    Michael Crawshaw. Multi-task learning with deep neural networks: A survey. CoRR, abs/2009.09796, 2020. 1, 2

    arXiv 2009

  3. [3]

    QLoRA: Efficient finetuning of quantized llms

    Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. QLoRA: Efficient finetuning of quantized llms. In Advances in Neural Information Processing Sys- tems, pages 10088–10115, 2023. 1

    work page 2023

  4. [4]

    BERT: Pre-training of deep bidirectional trans- formers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of deep bidirectional trans- formers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the As- sociation for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages 4171–4186, Minnea...

    work page 2019

  5. [5]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representa- tions, 2021. 2, 4

    work page 2021

  6. [6]

    Omnivore: A sin- gle model for many visual modalities

    Rohit Girdhar, Mannat Singh, Nikhila Ravi, Laurens van der Maaten, Armand Joulin, and Ishan Misra. Omnivore: A sin- gle model for many visual modalities. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16102–16112, 2022. 1, 2

    work page 2022

  7. [7]

    Parameter-efficient fine-tuning for large models: A comprehensive survey

    Zeyu Han, Chao Gao, Jinyang Liu, Jeff Zhang, and Sai Qian Zhang. Parameter-efficient fine-tuning for large models: A comprehensive survey. Transactions on Machine Learning Research, 2024. 1, 2

    work page 2024

  8. [8]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016. 2

    2016

Show all 43 references
  1. [9]

    Parameter-efficient transfer learning for NLP

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. Parameter-efficient transfer learning for NLP. In Proceedings of the 36th International Conference on Machine Learning, pages 2790–27...

  2. [10]

    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. InIn- ternational Conference on Learning Representations , 2022. 1, 2

    2022

  3. [11]

    Parameter-efficient Multi- task Fine-tuning for Transformers via Shared Hypernet- works

    Rabeeh Karimi Mahabadi, Sebastian Ruder, Mostafa De- hghani, and James Henderson. Parameter-efficient Multi- task Fine-tuning for Transformers via Shared Hypernet- works. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th In...

    2021

  4. [12]

    The kinetics human action video dataset

    Will Kay, Jo ˜ao Carreira, Karen Simonyan, Brian Zhang, Chloe Hillier, Sudheendra Vijayanarasimhan, Fabio Viola, Tim Green, Trevor Back, Paul Natsev, Mustafa Suleyman, and Andrew Zisserman. The kinetics human action video dataset. CoRR, abs/1705.06950, 2017. 2, 4

    2017 arXiv

  5. [13]

    Human action recognition and predic- tion: A survey

    Yu Kong and Yun Fu. Human action recognition and predic- tion: A survey. International Journal of Computer Vision , 130(5):1366–1401, 2022. 2

    2022

  6. [14]

    Poggio, and Thomas Serre

    Hildegard Kuehne, Hueihan Jhuang, Est ´ıbaliz Garrote, Tomaso A. Poggio, and Thomas Serre. HMDB: A large video database for human motion recognition. In IEEE International Conference on Computer Vision, ICCV 2011, Barcelona, Spain, November 6-13, 2011 , pages 2556–2563. IEEE C...

    2011

  7. [15]

    Efficient multi-domain learning by covariance normalization

    Yunsheng Li and Nuno Vasconcelos. Efficient multi-domain learning by covariance normalization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 1, 2

    2019

  8. [16]

    REPAIR: Removing repre- sentation bias by dataset resampling

    Yi Li and Nuno Vasconcelos. REPAIR: Removing repre- sentation bias by dataset resampling. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 2

    2019

  9. [17]

    RESOUND: To- wards Action Recognition Without Representation Bias

    Yingwei Li, Yi Li, and Nuno Vasconcelos. RESOUND: To- wards Action Recognition Without Representation Bias. In Computer Vision – ECCV 2018, pages 520–535, 2018. 2

    2018

  10. [18]

    Multi-dataset Training of Transformers for Robust Action Recognition

    Junwei Liang, Enwei Zhang, Jun Zhang, and Chunhua Shen. Multi-dataset Training of Transformers for Robust Action Recognition. Advances in Neural Information Processing Systems, 35:14475–14488, 2022. 1, 2

    2022

  11. [19]

    PolyViT: Co-training vision transformers on images, videos and audio

    Valerii Likhosherstov, Anurag Arnab, Krzysztof Marcin Choromanski, Mario Lucic, Yi Tay, and Mostafa Dehghani. PolyViT: Co-training vision transformers on images, videos and audio. Transactions on Machine Learning Research ,

  12. [20]

    Polyhistor: Parameter-Efficient Multi-Task Adaptation for Dense Vision Tasks

    Yen-Cheng Liu, Chih-Yao Ma, Junjiao Tian, Zijian He, and Zsolt Kira. Polyhistor: Parameter-Efficient Multi-Task Adaptation for Dense Vision Tasks. Advances in Neu- ral Information Processing Systems, 35:36889–36901, 2022. Polyhistor. 1

    2022

  13. [21]

    Decoupled weight de- cay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight de- cay regularization. In International Conference on Learning Representations, 2019. 4

    2019

  14. [22]

    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

  15. [23]

    Learning multiple visual domains with residual adapters

    Sylvestre-Alvise Rebuffi, Hakan Bilen, and Andrea Vedaldi. Learning multiple visual domains with residual adapters. In Advances in Neural Information Processing Systems , 2017. 1, 2

    2017

  16. [24]

    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), 2018. 1, 2

    2018

  17. [25]

    Imagenet-21k pretraining for the masses

    Tal Ridnik, Emanuel Ben-Baruch, Asaf Noy, and Lihi Zelnik-Manor. Imagenet-21k pretraining for the masses. In Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1) , 2021. 4

    2021

  18. [26]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10684–10695, 2022. 2

    2022

  19. [27]

    Johansen, Sergio Escalera, Kamal Nasrollahi, Thomas B

    Javier Selva, Anders S. Johansen, Sergio Escalera, Kamal Nasrollahi, Thomas B. Moeslund, and Albert Clap ´es. Video transformers: A survey. IEEE Transactions on Pattern Anal- ysis and Machine Intelligence , 45(11):12922–12943, 2023. 2

    2023

  20. [28]

    UCF101: A dataset of 101 human actions classes from videos in the wild

    Khurram Soomro, Amir Roshan Zamir, and Mubarak Shah. UCF101: A dataset of 101 human actions classes from videos in the wild. CoRR, abs/1212.0402, 2012. 2, 4

    2012 arXiv

  21. [29]

    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 (CVPR), pages 5227– 5237, 2022. 1

    2022

  22. [30]

    Antonio Torralba and Alexei A. Efros. Unbiased look at dataset bias. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1521–1528, 2011. ISSN: 1063-6919. 1

    2011

  23. [31]

    Vision transformers for action recognition: A survey

    Anwaar Ulhaq, Naveed Akhtar, Ganna Pogrebna, and Ajmal Mian. Vision transformers for action recognition: A survey. CoRR, abs/2209.05700, 2022. 2

    2022 arXiv

  24. [32]

    Multi-task learning for dense prediction tasks: A survey

    Simon Vandenhende, Stamatios Georgoulis, Wouter Van Gansbeke, Marc Proesmans, Dengxin Dai, and Luc Van Gool. Multi-task learning for dense prediction tasks: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(7):3614–3633, 2022. 1, 2

    2022

  25. [33]

    Interpretable image recognition by constructing transparent embedding space

    Jiaqi Wang, Huafeng Liu, Xinyue Wang, and Liping Jing. Interpretable image recognition by constructing transparent embedding space. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision (ICCV) , pages 895– 904, 2021. 4

    2021

  26. [34]

    Mimetics: To- wards understanding human actions out of context

    Philippe Weinzaepfel and Gr ´egory Rogez. Mimetics: To- wards understanding human actions out of context. Interna- tional Journal of Computer Vision, 129(5):1675–1690, 2021. 2

    2021

  27. [35]

    Large multimodal agents: A survey

    Junlin Xie, Zhihong Chen, Ruifei Zhang, Xiang Wan, and Guanbin Li. Large multimodal agents: A survey. CoRR, abs/2402.15116, 2024. 1, 2

    2024 arXiv

  28. [36]

    VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding

    Yi Xin, Junlong Du, Qiang Wang, Zhiwen Lin, and Ke Yan. VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding. In Proceedings of the AAAI Conference on Artificial Intelligence, pages 16085– 16093, 2024. Number: 14. 1, 2

    2024

  29. [37]

    Parameter-efficient fine- tuning for pre-trained vision models: A survey

    Yi Xin, Siqi Luo, Haodi Zhou, Junlong Du, Xiaohong Liu, Yue Fan, Qing Li, and Yuntao Du. Parameter-efficient fine- tuning for pre-trained vision models: A survey. CoRR, abs/2402.02242, 2024. 1, 2

    2024

  30. [38]

    A survey of efficient fine- tuning methods for Vision-Language Models — Prompt and Adapter

    Jialu Xing, Jianping Liu, Jian Wang, Lulu Sun, Xi Chen, Xunxun Gu, and Yingfei Wang. A survey of efficient fine- tuning methods for Vision-Language Models — Prompt and Adapter. Computers & Graphics, 119:103885, 2024. 1, 2

    2024

  31. [39]

    Peng Xu, Xiatian Zhu, and David A. Clifton. Multimodal Learning With Transformers: A Survey. IEEE Transac- tions on Pattern Analysis and Machine Intelligence, 45(10): 12113–12132, 2023. 1, 2

    2023

  32. [40]

    Tensors for Data Processing: Theory, Methods, and Applications

    Liu Yipeng, editor. Tensors for Data Processing: Theory, Methods, and Applications . Academic Press, 1st edition,

  33. [41]

    LLaMA-adapter: Efficient fine-tuning of large language models with zero- initialized attention

    Renrui Zhang, Jiaming Han, Chris Liu, Aojun Zhou, Pan Lu, Yu Qiao, Hongsheng Li, and Peng Gao. LLaMA-adapter: Efficient fine-tuning of large language models with zero- initialized attention. In The Twelfth International Conference on Learning Representations, 2024. 1

    2024

  34. [42]

    A survey on multi-task learning

    Yu Zhang and Qiang Yang. A survey on multi-task learning. IEEE Transactions on Knowledge and Data Engineering, 34 (12):5586–5609, 2022. 1, 2

    2022

  35. [43]

    Sim- ple multi-dataset detection

    Xingyi Zhou, Vladlen Koltun, and Philipp Kr¨ahenb¨uhl. Sim- ple multi-dataset detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 7571–7580, 2022. 1, 2

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.