Pith. sign in

REVIEW 4 major objections 4 minor 59 references

TADFormer : Task-Adaptive Dynamic Transformer for Efficient Multi-Task Learning

T0 review · 4 major / 4 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 TADFormer claims that input-conditioned task prompts and a dynamic task filter let a frozen Swin encoder outperform full fine-tuning on multi-task dense prediction while training up to about 8 times fewer parameters.

desk verdict The headline numbers are strong and the ablations are clean, but the TPC operator is underspecified for Swin's windowed attention, so the paper needs a clarifying revision before it can be fully trusted. read the letter →

arxiv 2501.04293 v2 pith:DY6LZREW submitted 2025-01-08 cs.CV

Seungmin Baek , Soyul Lee , Hayeon Jo , Hyesong Choi , Dongbo Min This is my paper · ORCID
classification cs.CV
keywords multi-tasklearningparameter-efficientfine-tuningtaskpromptingdynamicconvolutiondensepredictiontransformerlow-rankadaptationvisualprompttuning
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 a frozen pre-trained Transformer encoder can be adapted to several dense-prediction tasks at once with fewer trainable parameters than full fine-tuning and with higher accuracy, by making the adaptation depend on each input image's context rather than on static per-task parameters. The proposed TADFormer prepends a small set of trainable task prompts to the image patch tokens, reads the resulting task-attention map from the last self-attention block of each stage, and uses those maps to build task-adapted features. Those features pass through a low-rank module augmented by a dynamic task filter, a set of convolution kernels generated on the fly from global-average-pooled features, so the same module serves different tasks differently depending on the input. On the PASCAL-Context benchmark, the method reports higher overall multi-task accuracy than full fine-tuning and than recent PEFT baselines such as MTLoRA, while training roughly 4 to 8 times fewer parameters. If correct, this means input-conditioned task prompts can carry much of the task-specific adaptation that would otherwise require tuning most of the model.

What carries the argument

The load-bearing objects are the Task-Prompt Conditional (TPC) operator and the Dynamic Task Filter (DTF). The TPC operator takes the attention map $A_{TAM} \in \mathbb{R}^{H \times T \times N}$ between $T$ task prompts and $N$ patch tokens from the last MHSA of each stage and produces a task-adapted feature $f_i = f_{qkv} + S_{inv}(a_i \otimes \hat{f}_{qkv})$ for each task $i$; this is what injects task identity into the frozen encoder without training the encoder. The DTF sits between the low-rank down- and up-projections of the task-aware module, generates kernels $\theta_i = \phi(f_i W_{down})$ from global-average-pooled down-projected features, and applies channel-wise convolution so the output is $\tilde{F}_i = \Phi(f_i) + (\theta_i \odot (f_i W_{down})) W_{up}$, with FilterNorm stabilizing training. The two act in sequence: the TPC operator decouples per-task features from task-agnostic ones, and the DTF adapts those features to the specific input while a cross-task projection lets tasks exchange information; a gated skip connection then mixes the task-adapted feature with the block output before the task-specific decoder.

What would settle it

Run the rank-32 TADFormer on PASCAL-Context with the task attention map $A_{TAM}$ replaced by a fixed random map of the same shape, keeping all trained weights fixed; if the reported $\Delta m$ of $+3.63$ does not fall by roughly the $0.39$ points the ablation attributes to the TAM, then the TPC operator is not contributing through the attention it claims to compute.

Watch

Extended reading notes

Core claim

At rank 16, TADFormer reports $\Delta m = +2.44\%$ with 3.56M trainable parameters, against $+2.23\%$ with 30.06M for full fine-tuning, and at rank 64 it reports $+4.24\%$ with 7.38M parameters, against $+2.55\%$ with 8.34M for MTLoRA at the same rank. The claim is that this gain comes from two interacting mechanisms: the Task-Prompt Conditional (TPC) operator, which computes per-task attention maps between task prompts and image patch tokens inside the last multi-head self-attention of each stage and rescales patch features with them, and the Dynamic Task Filter (DTF), which generates channel-wise convolution kernels from global-average-pooled, down-projected task-adapted features, so that task-specific filtering adapts to the input context. Ablations attribute roughly $+1.65\,\Delta m$ to the DTF, $+0.49$ to the TPC operator plus task prompts, and $+0.39$ specifically to the task attention map, with the two components together adding $+2.33$ over a LoRA-only baseline. The paper also claims the design transfers to larger backbones and pretraining datasets and can be grafted onto adapter-based PEFT methods, improving their multi-task accuracy.

Load-bearing premise

The design assumes that the attention weights between the task prompts and image patch tokens, read from the last self-attention layer of each stage, still carry task-relevant context after passing through earlier frozen layers, and that global average pooling of down-projected features does not discard the spatial cues the dynamic filters need.

Editorial extensions

If this is right

  • At rank 64, TADFormer reaches $\Delta m = +4.24\%$ on PASCAL-Context with 7.38M trainable parameters, exceeding both full fine-tuning's $+2.23\%$ with 30.06M parameters and MTLoRA's $+2.55\%$ with 8.34M parameters.
  • The DTF alone contributes about $+1.65\,\Delta m$ over the LoRA-only baseline, and the TPC operator with task prompts contributes about $+0.49\,\Delta m$; the task attention map itself is worth about $0.39\,\Delta m$ at equal parameter counts.
  • Switching the backbone from Swin-T on ImageNet-1k to Swin-B or to ImageNet-22k pretraining increases the relative gain, indicating the method scales with model and data size rather than saturating.
  • Grafting the TPC-plus-DTF design onto adapter-based PEFT frameworks improves their multi-task accuracy, e.g., AdaptFormer with the design reaches $\Delta m = +2.69\%$ versus $-0.63\%$ for AdaptFormer alone at comparable parameter counts.

Reading between the lines

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

  • The paper does not test whether the task attention map would survive insertion of prompts at every stage rather than only the first; a natural follow-up is to measure whether prompt information is diluted by layer depth, which would directly probe the weakest assumption.
  • Because the DTF generates kernels from global-average-pooled features, a spatially sensitive variant that pools per region or uses multi-scale pooling might recover spatial cues the current GAP discards; the paper's own framing suggests this as a testable extension, but it does not investigate it.
  • The reported 0.39-point gap induced by the attention map is measured at equal trainable parameter counts, so it isolates the information content of the TAM; an adversarial test that corrupts or shuffles the TAM at inference would tell whether the model has learned to rely on it or has found a shortcut.
  • If the same recipe transfers to other frozen backbones that expose per-head attention maps, the mechanism could become a generic plug-in for multi-task PEFT beyond the Swin and HRNet combination used here; the paper only demonstrates Swin backbones.
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

4 major / 4 minor

Summary. This paper proposes TADFormer, a parameter-efficient fine-tuning method for multi-task dense prediction. It prepends task prompts to image patch tokens, uses LoRA-style task-shared modules in most Swin Transformer blocks, and introduces a Task-Prompt Conditional (TPC) operator plus a Dynamic Task Filter (DTF) in the last block of each stage. The TPC operator uses attention between task prompts and patch tokens to produce task-adapted features, while the DTF generates channel-wise convolutional filters conditioned on input features. Experiments on PASCAL-Context with four dense prediction tasks report higher Δm than MTLoRA and full fine-tuning with fewer trainable parameters, and the paper also provides ablations, decoder/backbone variants, and an adapter-based extension.

Significance. If the reported results are reproducible, TADFormer is a useful contribution: it combines prompt-based task adaptation with input-conditioned filters in a parameter-efficient multi-task framework, and the ablation study gives per-component evidence for the TPC operator and DTF. The paper provides full per-task tables, multiple rank settings, decoder/backbone variants, computational cost analysis, and attention-map visualizations. However, the central mechanism is underspecified for the Swin windowed-attention backbone, the main code and seed-level statistics are not provided, and the adapter extension contains severely degraded parallel configurations that are presented too positively. These issues currently prevent verification of the headline claim that TADFormer consistently outperforms existing PEFT methods.

major comments (4)
  1. [3.3.1 / Eq. (2) / 4.1] The TPC operator is underspecified for the Swin-T backbone. The text defines a global attention map A in R^{H x (T+N) x (T+N)} and extracts ATAM in R^{H x T x N} between task prompts and every patch token, but Swin's windowed MHSA only produces per-window attention matrices. If task prompts are simply prepended to the patch-token sequence, the prompts lie in one window and most patch tokens have no direct prompt attention, making Eq. (2) not evaluable as written. If a modified global attention or per-window prompt insertion is used, this is a significant architectural change that must be stated and reflected in Figure 4 and the equations. Relative position bias for prompt tokens also needs to be defined. This issue is load-bearing because Table 2 credits the task attention map with +0.39 Δm and the full TPC+DTF combination with +2.33 Δm over the baseline.
  2. [Supplementary Section 8, Table 4] The adapter extension is presented as evidence of generality, but the parallel configurations show a severe performance collapse. AdaptFormer parallel rows A2 and A4 have semantic segmentation mIoU 55.28 and 55.18 with Δm -10.54 and -11.06; TADFormer versions AO2 and AO4 improve to 61.41 and 60.37 with Δm -5.02 and -6.25, still far below the MTL full-fine-tuning baseline (+2.23, semseg 67.56). The text's conclusion that TADFormer yields an overall enhancement in parallel configurations is true only in the narrow sense of making a badly degraded configuration less degraded. Please diagnose this failure or remove the parallel-configuration claim; as it stands, the claim that TADFormer is compatible with adapter-based frameworks is not supported.
  3. [Table 1 and Section 4.1] All main results are single-run numbers without error bars, seed counts, or released main-experiment code. The headline differences (e.g., TADFormer r=16 Δm +2.44 vs MTLoRA r=16 +1.35, and the reported 1.2-1.7 Δm gains at r=32/64) are small enough that seed variance in multi-task dense prediction could affect the ranking. Please report mean and standard deviation over at least three seeds for the main comparisons and make the main training/evaluation code available. Without this, the central claim of consistent superiority over MTLoRA cannot be verified.
  4. [3.4.1 / Eqs. (3)-(4)] The DTF is motivated by the need to capture input context, but its parameter generation network uses global average pooling of the down-projected feature, which discards spatial structure before generating the channel-wise filters. The paper does not explain how spatially global statistics preserve the fine-grained, input-dependent features that the method claims to extract, and no analysis is provided beyond the single ablation row and the Grad-CAM figure. A mechanism analysis or an ablation comparing GAP with spatial pooling would be needed to support the design claim.
minor comments (4)
  1. [3.5 / Eq. (5)] With g initialized to zero, sigmoid(g)=0.5, so Eq. (5) begins by averaging fi and the frozen-path output rather than starting from the frozen path; please clarify whether this is intended and, if so, motivate the equal-weight initialization.
  2. [4.2 / References] The Adapter baseline is cited as [17], but [17] is the 'Towards a unified view' paper; the original adapter citation should be checked and corrected.
  3. [Figure 2] The label 'TCP-operator' in Figure 2(b) is inconsistent with the 'TPC-operator' used throughout the text; please fix the acronym.
  4. [3.3.1] The notation fqkv is introduced as the output of the 'QKV module' without specifying whether it refers to the value features, the projected output, or the entire QKV concatenation; please define it precisely to make Eq. (2) unambiguous.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are held-out empirical measurements, and the self-citations are non-load-bearing.

full rationale

The paper makes no first-principles prediction whose value is forced by its own definitions. The TPC operator is defined by Eq. 2 from the MHSA attention map and the DTF by Eqs. 3-4, but neither equation is fitted to the evaluation metric Δm; Table 1 and Table 2 report measured test-set numbers across configurations. The task weights are borrowed from [44] and the training setup from [1], but these are external inputs to the experiment, not quantities that analytically determine the reported Δm values. The self-citations ([8], [9], [10], [29], [30], [31]) appear in related work and as inspiration in Section 3.3; none is invoked as a theorem, a uniqueness claim, or the sole justification for a central result. The Swin windowed-attention underspecification raised by the skeptical review is a reproducibility or correctness concern, not a circular reduction: it does not make Eq. 2 equivalent to the headline improvement. No quoted step exhibits a fitted parameter renamed as a prediction or a definition that reduces to the target result, so no circularity is established under the required standard.

Assumptions & free parameters 4 free parameters · 4 assumptions · 2 invented entities

The central claim rests on a handful of design choices (rank, DTF kernel size, gating initialization, borrowed task weights) and on the assumption that attention maps and GAP-based filter generation preserve task-relevant context. No new physical entities are introduced; the DTF and TPC operator are architectural components whose effectiveness is supported only by the paper's own experiments.

free parameters (4)
  • rank r of TS/TA low-rank projections = 16, 32, 64
    Capacity swept in all experiments; reported gains depend on it (e.g., delta-m +2.44 at r=16 vs +4.24 at r=64 in Table 1).
  • DTF convolution kernel size k = not reported
    Controls the number of filter parameters r times r times k squared in Eqs. 3-4; the value is not stated in the main text, so a re-implementation must guess it.
  • gating parameter g initial value = 0
    Initialized to zero with sigmoid in Eq. 5 to stabilize early training; a hand-chosen schedule-free design choice.
  • task loss weights w_i = from MTI-Net [44]
    Used in Eq. 7 for the MTL loss; borrowed from prior work rather than learned, but the central results depend on this particular weighting.
assumptions (4)
  • domain assumption Attention weights between task prompts and patch tokens in the last MHSA of each stage are a usable task-relevant map.
    Section 3.3.1 builds the TPC operator on the task attention map; no analysis is given of how prompt information survives earlier stages.
  • domain assumption Channel-wise convolution filters generated by global average pooling of down-projected features capture task-specific input context.
    Section 3.4.1 and Figure 5; the paper cites dynamic filter networks but provides no justification that GAP preserves the needed details.
  • domain assumption Comparing against MTLoRA under 'the same experimental setup as [1]' is a fair baseline.
    Section 4.1; no code or hyperparameter listing is given, so the comparison rests on this assertion.
  • domain assumption One benchmark (PASCAL-Context with four tasks) supports conclusions about MTL PEFT generally.
    Section 4.1; supplementary adds other backbones and decoders but still on the same dataset.
invented entities (2)
  • Dynamic Task Filter (DTF)
    purpose: Generates input-conditioned channel-wise convolution weights inside the TA-module (Eqs. 3-4).
    Evidence is only the paper's own benchmark tables and Grad-CAM; no external prediction or code is provided to verify independently.
  • Task-Prompt Conditional (TPC) operator
    purpose: Creates task-adapted features by multiplying the task attention map with reshaped QKV features (Eq. 2).
    Its effectiveness rests on the paper's ablation and visualization only, with no external verification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TADFormer : Task-Adaptive Dynamic Transformer for Efficient Multi-Task Learning." pith.science (2026). https://pith.science/paper/DY6LZREW

@misc{pith2026250104293,
  author       = {Pith},
  title        = {Pith review of: TADFormer : Task-Adaptive Dynamic Transformer for Efficient Multi-Task Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DY6LZREW}},
  note         = {Machine review of arXiv:2501.04293}
}
read the original abstract

Transfer learning paradigm has driven substantial advancements in various vision tasks. However, as state-of-the-art models continue to grow, classical full fine-tuning often becomes computationally impractical, particularly in multi-task learning (MTL) setup where training complexity increases proportional to the number of tasks. Consequently, recent studies have explored Parameter-Efficient Fine-Tuning (PEFT) for MTL architectures. Despite some progress, these approaches still exhibit limitations in capturing fine-grained, task-specific features that are crucial to MTL. In this paper, we introduce Task-Adaptive Dynamic transFormer, termed TADFormer, a novel PEFT framework that performs task-aware feature adaptation in the fine-grained manner by dynamically considering task-specific input contexts. TADFormer proposes the parameter-efficient prompting for task adaptation and the Dynamic Task Filter (DTF) to capture task information conditioned on input contexts. Experiments on the PASCAL-Context benchmark demonstrate that the proposed method achieves higher accuracy in dense scene understanding tasks, while reducing the number of trainable parameters by up to 8.4 times when compared to full fine-tuning of MTL models. TADFormer also demonstrates superior parameter efficiency and accuracy compared to recent PEFT methods.

Figures

Figures reproduced from arXiv: 2501.04293 by the authors.

Figure 1
Figure 1. Performance comparison between TADFormer and base [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. b, task-adapted features are generated with the help of task attention maps that represent task-specific attributes for each task. They are then fed into a Task-Aware Module (TA-Module) augmented by Dynamic Task Filter (DTF). The DTF adaptively captures fine-grained features by lever￾aging the context information of input samples through dy￾namic convolution operations. Furthermore, the distinct Pretrained Module Sh… view at source ↗
Figure 3
Figure 3. Comparison of Grad-CAM [40] from MTLoRA [1] and TADFormer: (from top to bottom) input images, MTLoRA, and TADFormer. This demonstrates that TADFormer is capable of ex￾tracting fine-grained features that capture the input contexts more precisely, thanks to DTF. models is challenging in that unique attributes of each task should be extracted through pre-trained models only with small amount of learnable parts. MTLoRA … view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Overview of the proposed TADFormer: (a) The encoder takes as inputs image patch tokens with task prompts prepended. Here, [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: DTF architecture: The down-projected features of chan [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Patch merging module of TADFormer: Leveraging [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: Performance of TADFormer when employing the Swin [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: (a) illustrates the performance difference when using [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: Efficiency of TADFormer with different number of tasks: The experiments were run at the rank r = 32. 8. TADFormer on Adapter-Based Methods Experimental Setup. To analyze the extensibility of TAD￾Former into other PEFT methods, we experimented TAD￾Former with adapter-b…
Figure 11
Figure 11. Figure 11: Adapter configuration: Both sequential and parallel [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: Overview of Adapter + TADFormer in parallel adapter [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 41 canonical work pages

  1. [1]

    Mtlora: Low-rank adaptation approach for efficient multi-task learn- ing

    Ahmed Agiza, Marina Neseem, and Sherief Reda. Mtlora: Low-rank adaptation approach for efficient multi-task learn- ing. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 16196–16205,

    work page

  2. [2]

    Adversarial training for multi-context joint entity and relation extraction

    Giannis Bekoulis, Johannes Deleu, Thomas Demeester, and Chris Develder. Adversarial training for multi- context joint entity and relation extraction. arXiv preprint arXiv:1808.06876, 2018. 3

    work page Pith review arXiv 2018

  3. [3]

    Mult: An end-to-end multitask learning transformer

    Deblina Bhattacharjee, Tong Zhang, Sabine S ¨usstrunk, and Mathieu Salzmann. Mult: An end-to-end multitask learning transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12031– 12041, 2022. 1, 3

    work page 2022

  4. [4]

    Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D. Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Lan- guage models are few-shot learners. In Advances in Neural Information Processing Systems, 2020. 1

    work page 2020

  5. [5]

    End-to- end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to- end object detection with transformers. In European confer- ence on computer vision, pages 213–229. Springer, 2020. 3

    2020

  6. [6]

    Adapter configuration: Both sequential and parallel configurations are possible in adapter-based PEFT framework [7]

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian … Task attention map Task-agnostic Feature Task-Adapted Feature … Pretrained Module down up Task-agnostic Feature + Task-Shared Module … Pretrained Module DTF down up + … Pretrained Module DTF down up + TA-Module (Projection) TA-Module (Linear 1,2) Frozen Tuned Task-Agnostic Feature route Task-Specif...

    work page 2018

  7. [7]

    Adaptformer: Adapting vision transformers for scalable visual recogni- tion

    Shijie Chen, Chongjian Ge, Zhanhong Tong, Jianmin Wang, Yang Song, Jianbo Wang, and Ping Luo. Adaptformer: Adapting vision transformers for scalable visual recogni- tion. In Advances in Neural Information Processing Systems, pages 16664–16678, 2022. 3, 8, 2, 4

    work page 2022

  8. [8]

    Improving Generative Pre-Training: An In-depth Study of Masked Image Modeling and Denoising Models

    Hyesong Choi, Daeun Kim, Sungmin Cha, Kwang Moo Yi, and Dongbo Min. Improving generative pre-training: An in- depth study of masked image modeling and denoising mod- els. arXiv preprint arXiv:2412.19104, 2024. 3

    work page Pith review arXiv 2024

Show all 59 references
  1. [9]

    Emerging property of masked token for effective pre-training

    Hyesong Choi, Hunsang Lee, Seyoung Joung, Hyejin Park, Jiyeong Kim, and Dongbo Min. Emerging property of masked token for effective pre-training. arXiv preprint arXiv:2404.08330, 2024

    2024 arXiv

  2. [10]

    Salience-based adaptive masking: revisit- ing token dynamics for enhanced pre-training

    Hyesong Choi, Hyejin Park, Kwang Moo Yi, Sungmin Cha, and Dongbo Min. Salience-based adaptive masking: revisit- ing token dynamics for enhanced pre-training. In European Conference on Computer Vision , pages 343–359. Springer,

  3. [11]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009. 6

    2009

  4. [12]

    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 Annual Conference of the North American Chapter of the Association for Computational Linguistics, 2018. 1

    2018

  5. [13]

    An image is worth 16x16 words: Transformers for image recognition at scale, 2021

    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 ...

    2021

  6. [14]

    Llama-adapter v2: Parameter-efficient vi- sual instruction model

    Peng Gao, Jiaming Han, Renrui Zhang, Ziyi Lin, Shijie Geng, Aojun Zhou, Wei Zhang, Pan Lu, Conghui He, Xi- angyu Yue, et al. Llama-adapter v2: Parameter-efficient vi- sual instruction model. arXiv preprint arXiv:2304.15010 ,

    arXiv

  7. [15]

    Yihang Gao, Jun Ma, Mingbo Zhao, Wei Liu, and Alan L. Yuille. Nddr-cnn: Layerwise feature fusing in multi-task cnns by neural discriminative dimensionality reduction. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR), 2019. 3

    2019

  8. [16]

    Sensitivity-aware visual parameter-efficient fine-tuning

    Haotao He, Jianfei Cai, Jing Zhang, Dacheng Tao, and Bo- han Zhuang. Sensitivity-aware visual parameter-efficient fine-tuning. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 11825–11835, 2023. 3

    2023

  9. [17]

    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. 6, 7

    2021 arXiv

  10. [18]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollar, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2022. 1

    2022

  11. [19]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll´ar, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16000– 16009, 2022. 3

    2022

  12. [20]

    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 International Conference on Machine Learning, pages 2790–2799. PMLR, 2019. 1, 2, 3

    2019

  13. [21]

    Hu et al

    Edward J. Hu et al. Lora: Low-rank adaptation of large lan- guage models. arXiv preprint arXiv:2106.09685, 2021. 1, 2, 3, 4, 6, 7, 8

    2021 arXiv

  14. [22]

    Llm-adapters: An adapter family for parameter- efficient fine-tuning of large language models.arXiv preprint arXiv:2304.01933, 2023

    Zhiqiang Hu, Yihuai Lan, Lei Wang, Wanyu Xu, Ee- Peng Lim, Roy Ka-Wei Lee, Lidong Bing, and Soujanya Poria. Llm-adapters: An adapter family for parameter- efficient fine-tuning of large language models.arXiv preprint arXiv:2304.01933, 2023. 1

    2023 arXiv

  15. [23]

    Going beyond multi-task dense pre- diction with synergy embedding models

    Huimin Huang et al. Going beyond multi-task dense pre- diction with synergy embedding models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, page to appear, 2024. 1, 2, 3

    2024

  16. [24]

    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 European Conference on Computer Vision, pages 709–727. Springer, 2022. 3, 4, 6, 7

    2022

  17. [25]

    Dynamic filter networks

    Xu Jia, Bert De Brabandere, Tinne Tuytelaars, and Luc V Gool. Dynamic filter networks. Advances in neural informa- tion processing systems, 29, 2016. 5

    2016

  18. [26]

    Compacter: Efficient low-rank hypercomplex adapter layers

    Rabeeh Karimi Mahabadi, James Henderson, and Sebastian Ruder. Compacter: Efficient low-rank hypercomplex adapter layers. In Advances in Neural Information Processing Sys- tems, pages 1022–1035, 2021. 6, 7

    2021

  19. [27]

    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. 1, 6, 7

    2021 arXiv

  20. [28]

    Multi-task learning using uncertainty to weigh losses for scene geome- try and semantics

    Alex Kendall, Yarin Gal, and Roberto Cipolla. Multi-task learning using uncertainty to weigh losses for scene geome- try and semantics. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 7482– 7491, 2018. 3

    2018

  21. [29]

    Sequen- tial cross attention based multi-task learning

    Sunkyung Kim, Hyesong Choi, and Dongbo Min. Sequen- tial cross attention based multi-task learning. In 2022 IEEE International Conference on Image Processing (ICIP), pages 2311–2315. IEEE, 2022. 3

    2022

  22. [30]

    Knn local attention for image restoration

    Hunsang Lee, Hyesong Choi, Kwanghoon Sohn, and Dongbo Min. Knn local attention for image restoration. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 2139–2149, 2022. 3

    2022

  23. [31]

    Cross-scale knn image transformer for image restoration

    Hunsang Lee, Hyesong Choi, Kwanghoon Sohn, and Dongbo Min. Cross-scale knn image transformer for image restoration. IEEE Access, 11:13013–13027, 2023. 3

    2023

  24. [32]

    Conditional adapters: Parameter-efficient transfer learning with fast in- ference

    Tao Lei, Junlin Bai, Siddhartha Brahma, Joshua Ainslie, Kenton Lee, Yi Zhou, Ming-Wei Chang, et al. Conditional adapters: Parameter-efficient transfer learning with fast in- ference. In Advances in Neural Information Processing Sys- tems, pages 8152–8172, 2023. 3

    2023

  25. [33]

    The power of scale for parameter-efficient prompt tuning.arXiv preprint arXiv:2104.08691, 2021

    Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning.arXiv preprint arXiv:2104.08691, 2021. 1, 3

    2021 arXiv

  26. [34]

    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. 1, 3

    2021 arXiv

  27. [35]

    Polyhistor: Parameter-efficient multi-task adap- tation for dense vision tasks

    Yen-Cheng Liu, Chih-Yao Ma, Junjiao Tian, Zijian He, and Zsolt Kira. Polyhistor: Parameter-efficient multi-task adap- tation for dense vision tasks. In Advances in Neural Infor- mation Processing Systems, pages 36889–36901, 2022. 1, 6, 7

    2022

  28. [36]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 4, 5, 7

    2021

  29. [37]

    Swin trans- former: Hierarchical vision transformer using shifted win- dows, 2021

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin trans- former: Hierarchical vision transformer using shifted win- dows, 2021. 3

    2021

  30. [38]

    Swin transformer v2: Scaling up capacity and resolution

    Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, et al. Swin transformer v2: Scaling up capacity and resolution. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 12009–12019, 2022

    2022

  31. [39]

    Vi- sion transformers for dense prediction

    Ren ´e Ranftl, Alexey Bochkovskiy, and Vladlen Koltun. Vi- sion transformers for dense prediction. In Proceedings of the IEEE/CVF international conference on computer vision, pages 12179–12188, 2021. 3

    2021

  32. [40]

    Grad-cam: Visual explanations from deep networks via gradient-based localization

    Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-cam: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE in- ternational conference on computer vision , pages 618–626,

  33. [41]

    Multi-task learning as multi-objective optimization

    Ozan Sener and Vladlen Koltun. Multi-task learning as multi-objective optimization. In Advances in Neural Infor- mation Processing Systems, 2018. 3

    2018

  34. [42]

    Deep high-resolution representation learning for human pose es- timation

    Ke Sun, Bin Xiao, Dong Liu, and Jingdong Wang. Deep high-resolution representation learning for human pose es- timation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5693–5703,

  35. [43]

    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,

  36. [44]

    Mti-net: Multi-scale task interaction networks for multi-task learning

    Simon Vandenhende, Stamatios Georgoulis, and Luc Van Gool. Mti-net: Multi-scale task interaction networks for multi-task learning. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16, pages 527–543. Springer, 2020. 6

    2020

  37. [45]

    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, 2021. 2, 3

    2021

  38. [46]

    Segformer: Simple and efficient design for semantic segmentation with transform- ers

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M Alvarez, and Ping Luo. Segformer: Simple and efficient design for semantic segmentation with transform- ers. Advances in neural information processing systems, 34: 12077–12090, 2021. 3

    2021

  39. [47]

    Segformer: Simple and efficient design for semantic segmentation with transform- ers

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M Alvarez, and Ping Luo. Segformer: Simple and efficient design for semantic segmentation with transform- ers. Advances in neural information processing systems, 34: 12077–12090, 2021. 1, 3

    2021

  40. [48]

    Simmim: A simple framework for masked image modeling

    Zhenda Xie, Zheng Zhang, Yue Cao, Yutong Lin, Jianmin Bao, Zhuliang Yao, Qi Dai, and Han Hu. Simmim: A simple framework for masked image modeling. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022. 1, 3

    2022

  41. [49]

    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,

  42. [50]

    Multi-task dense prediction via mixture of low-rank experts

    Yutong Yang, Peng-Tao Jiang, Qibin Hou, Hao Zhang, Jinhui Chen, and Bing Li. Multi-task dense prediction via mixture of low-rank experts. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 27927–27937, 2024. 1, 3, 5

    2024

  43. [51]

    Taskprompter: Spatial-channel multi-task prompting for dense scene understanding

    Hanrong Ye and Dan Xu. Taskprompter: Spatial-channel multi-task prompting for dense scene understanding. In The Eleventh International Conference on Learning Representa- tions, 2022. 4

    2022

  44. [52]

    Taskexpert: Dynamically assem- bling multi-task representations with memorial mixture-of- experts

    Hanrong Ye and Dan Xu. Taskexpert: Dynamically assem- bling multi-task representations with memorial mixture-of- experts. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, 2023. 1, 2, 3

    2023

  45. [53]

    Masked image modeling with denoising contrast

    Kun Yi, Yixiao Ge, Xiaotong Li, Shusheng Yang, Dian Li, Jianping Wu, Ying Shan, and Xiaohu Qie. Masked image modeling with denoising contrast. arXiv preprint arXiv:2205.09616, 2022. 3

    2022 arXiv

  46. [54]

    1% vs 100%: Parameter-efficient low rank adapter for dense predictions

    Dongqi Yin, Yi Yang, Zilong Wang, Hao Yu, Ke Wei, and Xia Sun. 1% vs 100%: Parameter-efficient low rank adapter for dense predictions. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 20116–20126, 2023. 3

    2023

  47. [55]

    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,

  48. [56]

    Zamir, Alexander Sax, William Shen, Leonidas J

    Amir R. Zamir, Alexander Sax, William Shen, Leonidas J. Guibas, Jitendra Malik, and Silvio Savarese. Taskonomy: Disentangling task transfer learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), 2018. 8

    2018

  49. [57]

    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, 2021. 2

    2021

  50. [58]

    Convolution meets lora: Parameter efficient finetuning for segment anything model

    Zihan Zhong, Zhiqiang Tang, Tong He, Haoyang Fang, and Chun Yuan. Convolution meets lora: Parameter efficient finetuning for segment anything model. In Proceedings of the International Conference on Learning Representations ,

  51. [59]

    Decoupled dynamic filter networks

    Jingkai Zhou, Varun Jampani, Zhixiong Pi, Qiong Liu, and Ming-Hsuan Yang. Decoupled dynamic filter networks. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 6647–6656, 2021. 5

    2021

Pith tools

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