Pith. sign in

REVIEW 3 major objections 6 minor 75 references

Unlocking Tuning-Free Few-Shot Adaptability in Visual Foundation Models by Recycling Pre-Tuned LoRAs

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A visual foundation model equipped with a meta-LoRA distilled from pre-tuned LoRAs can solve new few-shot tasks in a single forward pass, without fine-tuning.

desk verdict A genuinely new tuning-free few-shot pipeline built from recycled LoRAs, with a coherent method and useful ablations, but the empirical claims need more rigorous validation before I'd trust the headline numbers. read the letter →

arxiv 2412.02220 v1 pith:TBD5KIMF submitted 2024-12-03 cs.CV cs.AIcs.LG

Zixuan Hu , Yongxian Wei , Li Shen , Chun Yuan , Dacheng Tao This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords LoRARecyclefew-shotlearningmeta-learningtuning-freeadaptationdata-freemodelinversionvisualfoundationmodelstokenpruning
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

Visual foundation models today adapt to a new task only by fine-tuning, which needs enough labeled data and adds latency; large language models, by contrast, adapt from a few examples in context. The paper claims this gap can be closed by recycling the many task-specific LoRA adapters that the fine-tuning ecosystem has already produced, without ever touching the private data those LoRAs were trained on. It distills a single meta-LoRA from surrogate images generated by inverting the pre-tuned LoRAs, and the meta-LoRA gives the frozen vision model the ability to solve an unseen N-way K-shot classification task in one forward pass. On five-way one-shot in-domain benchmarks, the method reports an average 6.27% gain over comparison methods while needing no test-time parameter updates.

What carries the argument

The load-bearing machinery is the meta-LoRA plus the two procedures that feed it. LoRA Inversion turns a pre-tuned LoRA into surrogate image data by minimizing classification loss and a batch-normalization statistics prior; the prior borrows statistics from an open-source convolutional network because the vision transformer has no batch norm. The meta-learning objective teaches the meta-LoRA to make the same query-set predictions as each pre-tuned teacher after only a support-set prototype computation, which is what makes adaptation parameter-free at test time. Cross-task interpolation mixes classes from different LoRAs to densify the task distribution, and a token-pruning step uses the class token's attention weights to discard background tokens during inversion and then reuses the resulting mask to meta-train on sparse tokens, cutting compute and reducing noise.

What would settle it

On CIFAR-FS the paper reports 91.21% with original training data versus 89.69% with generated surrogate data for 5-way 1-shot; a falsifying result would be a dataset where this gap grows to the size of the method's reported gains, because then the surrogate channel rather than the meta-learning objective would be carrying the performance. The same experiment with the borrowed statistics replaced by the true training-data statistics would isolate whether the naturalness prior is the load-bearing part.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a task-agnostic adapter can be meta-learned from the low-rank residuals of many task-specific adapters. Each pre-tuned LoRA is treated as a teacher: the method optimizes random Gaussian noise into synthetic images that the teacher classifies confidently (LoRA Inversion, Eq. (1)), then splits those images into support and query sets. A meta-LoRA is trained so that, when its nearest-class-center predictions on the query set diverge from the teacher's predictions, the divergence is minimized (Eq. (3)); because the inner adaptation is just a distance-to-centroid computation, no second-order gradients are needed. After training across many LoRAs and interpolated cross-LoRA tasks, the meta-LoRA is added to the frozen foundation model, and a new few-shot task is handled by embedding the support examples, forming class centers, and classifying the query examples in the same forward pass. The paper reports this works both when the recycled LoRAs come from the same domain as the test task and when they come from different domains.

Load-bearing premise

The load-bearing assumption is that synthetic images generated by reversing a LoRA, using statistics borrowed from a different image model to make them look natural, keep enough of the original private data's class-discriminating structure that a meta-LoRA trained on them transfers to real, unseen classes.

Editorial extensions

If this is right

  • A visual foundation model equipped with the meta-LoRA can take on a new few-shot task without any gradient updates, so deployment latency and memory for adaptation drop to roughly the cost of one forward pass.
  • LoRAs uploaded to public repositories can be reused for new tasks even when their original training data is private, reducing the data exposure in the fine-tuning ecosystem.
  • The method extends beyond same-domain reuse: LoRAs from general image datasets give measurable gains on medical and satellite-imagery few-shot tasks.
  • The double-efficient mechanism means meta-training and inversion costs fall as sparsity increases, with reported CUB meta-training throughput up to roughly 3x faster at 75% token sparsity while accuracy is maintained.
  • Recycling works across LoRA ranks, so the meta-LoRA does not depend on all source adapters sharing one architecture.

Reading between the lines

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

  • Editorial inference: the same distillation recipe should transfer to other modular adapters, such as adapters or prefix modules, provided their inversion surfaces are smooth enough to yield discriminative surrogate data; nothing in the argument is specific to the low-rank parameterization.
  • Editorial inference: the attention-derived mask that marks foreground tokens during inversion is itself a saliency signal; it could be reused as a data augmentation prior or as a way to audit what each LoRA has learned, independent of the meta-training step.
  • Editorial inference: if surrogate data is the only channel between private training sets and the meta-LoRA, then the method inherits the biases and failure modes of the LoRAs it recycles; a poisoned or adversarially tuned LoRA could plant a pattern into the synthetic images that the meta-LoRA cannot distinguish from a real class signal.
  • Editorial inference: because the method only needs a frozen backbone and a set of LoRAs, it is a plausible plug-in for video or multimodal foundation models where LoRA-style adapters are also common, though the inversion step would need to be reworked for non-image inputs.
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 / 6 minor

Summary. The paper proposes LoRA Recycle, a framework that reuses a collection of pre-tuned LoRAs without access to their original training data, in order to give a visual foundation model (VFM) tuning-free few-shot adaptability. Surrogate images are generated by inverting each pre-tuned LoRA through Eq. (1), and a single meta-LoRA is distilled from these surrogate data via the meta-learning objective in Eqs. (3) and (4). At meta-test time the VFM plus the meta-LoRA performs N-way K-shot classification in one forward pass using nearest-centroid predictions, with no parameter updates. The paper also introduces a double-efficient mechanism that prunes tokens during inversion and then trains the meta-LoRA only on the remaining tokens. Experiments cover in-domain and cross-domain few-shot benchmarks with ViT-B/16 and ViT-B/32, plus additional results on Meta-Dataset, zero-shot classification, different LoRA ranks, and other ViT backbones.

Significance. If the central claim holds, LoRA Recycle would be a practically valuable bridge between the LoRA-market ecosystem and data-free meta-learning: it would let users convert a pool of task-specific LoRAs into a single lightweight module that adapts to new few-shot tasks without fine-tuning, addressing privacy and latency constraints. The paper has several concrete strengths: it meta-trains only 0.14M parameters (about 0.1% of the VFM), uses a first-order objective that avoids Hessian computations, reports substantial compute savings from token pruning (up to roughly 3x meta-training throughput and large memory reductions in Table 5), and evaluates across a broad set of benchmarks including cross-domain transfer and Meta-Dataset. The main uncertainty is empirical rather than conceptual: the framework's correctness rests on whether surrogate data generated by Eq. (1) preserve the class-discriminative structure of the original private data well enough to transfer to real novel-class images.

major comments (3)
  1. [Sec. 4.1-4.2, Table 9] The load-bearing premise of the paper is that surrogate data generated by inverting pre-tuned LoRAs, Eq. (1), preserve enough class-discriminative structure that a meta-LoRA trained only on these data transfers to real novel-class images. The only direct evidence for this is Table 9, which is limited to CIFAR-FS in the in-domain setting, reports no error bars or confidence intervals, and leaves the cross-domain transfer scenario untested. The training signal is also partly self-referential: the teacher LoRA defines both the synthetic data (Eq. (1)) and the targets used to train the student (the KL term in Eq. (3a)), so the student could in principle learn inversion-specific artifacts rather than transferable features. I request three additions: (i) a cross-domain version of Table 9 that meta-trains on original data from the four source domains and evaluates on the same cross-domain benchmarks as Table 3; (ii) confidence intervals or standard errors over task samples and random seeds for at least Tables 2, 3, and 9; and (iii) an auxiliary check of teacher-transfer fidelity, such as comparing the accuracy of h_T followed by f_{deltaW_T} on generated query images versus real query images, to show that teacher predictions on synthetic data are not relying on artifacts.
  2. [Sec. 4.1, Eq. (2)] The naturalness prior R_BN in Eq. (2) is the only real-data anchor in the inversion process, yet it borrows batch-normalization statistics from an open-source ResNet50, an architecture that is not used in the CLIP ViT. The paper states this borrowing in Sec. 4.1 but provides no evidence that ResNet50 feature statistics are compatible with the CLIP ViT's feature maps. Because this prior directly shapes the surrogate data, the compatibility assumption is load-bearing. I recommend validating it by varying the source of the BN statistics, or by reporting the distance between generated and real feature statistics computed in the CLIP ViT feature space, and by showing that the inversion quality is not sensitive to the choice of the prior backbone.
  3. [Sec. 4.3, Fig. 3, Table 6] The double-efficient mechanism creates a train/test discrepancy: meta-training consumes masked synthetic images with only a subset of tokens, while meta-testing uses full real images with all tokens. Table 6 shows that the background classification loss is flat during inversion, but this does not establish that background tokens are pure noise in real test images, nor that the attention-derived mask on synthetic data transfers to real images. Since masking is part of the final pipeline and is reported to improve accuracy, I ask for an ablation that trains on full synthetic images under matched computational budget, and, if feasible, a test-time evaluation that applies the learned mask or attention mask to real support and query images. This would clarify whether the masking gains in Tables 2 and 3 come from noise reduction that generalizes or from a distributional shortcut that happens to help on synthetic data.
minor comments (6)
  1. [Section 1] The phrase "tuning-fee few-shot adaptation" should read "tuning-free few-shot adaptation."
  2. [Tables 2 and 3] The baseline name "CMAL" is inconsistent with the text and Appendix D, where the method is called "CAML"; please unify the spelling.
  3. [Tables 2 and 3] The superscript and subscript gain annotations are hard to parse, and the abstract's "average 6.27% improvement" is not transparently derived from the tables. Please state explicitly which LoRA Recycle configuration is the final method and how the average gain is computed.
  4. [Table 4] The first two numeric columns in Table 4 are accuracy values, but the column headers do not say so; please add "5-way 1-shot Acc" and "5-way 5-shot Acc" or otherwise clarify the table layout.
  5. [Eq. (3a)] The KL direction in Eq. (3a) is not stated explicitly. Please clarify that the divergence is taken between the student's predicted distribution and the teacher's output distribution, and specify the convention used.
  6. [Appendix D] There is a typo in the opening sentence: "implementation details for the baselines used in our paper.." contains a doubled period.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the meta-LoRA is trained on teacher-generated surrogate data but evaluated on disjoint real-class tasks, so the central claim does not reduce to its inputs.

full rationale

LoRA Recycle's pipeline is data-free meta-learning: Eq. (1) generates surrogate images by inverting each pre-tuned LoRA, Eq. (3a) distills a meta-LoRA by matching the teacher LoRA's predictions on those synthetic query images, and test-time inference is a fixed nearest-centroid rule on real support/query images using the meta-LoRA. The teacher thus defines the meta-training signal, but the paper's claims are about an external evaluation on real images from label-disjoint meta-test tasks (Sec. 5.1, Sec. 5.2). No parameter of the meta-LoRA is fitted to the test support or query labels, and the reported accuracies are not forced by Eq. (1) or Eq. (3a) by construction. The only direct link between synthetic and real data is the empirical transfer, and Table 9 explicitly ablates it by comparing meta-training on generated data vs. original training data (91.21 vs. 89.69 in 1-shot, 98.93 vs. 97.05 in 5-shot), which is the appropriate check rather than a circular step. Self-citations ([28]-[30], [66]-[67]) appear only in related-work and appendix discussions of data-free meta-learning and are not invoked as the justification for the central result. The ResNet50 batch-normalization prior in Eq. (2), borrowed from external work, is an assumption about natural-image statistics, not a self-referential definition. No equation is defined in terms of the quantity it is claimed to predict, and no fitted parameter is renamed as a prediction. Therefore no circular step is present.

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

No new physical or ontological entities are postulated. The meta-LoRA is a trainable parameter set, not a new entity with independent evidence requirements. The central claim rests on the free parameters and domain assumptions listed above.

free parameters (5)
  • alpha_R = 0.01
    Coefficient balancing cross-entropy and naturalness prior in Eq. (1); tuned via sensitivity analysis in Table 18, with values 0.1, 0.01, 0.001 explored.
  • LoRA inversion learning rate = 0.25
    Learning rate for optimizing surrogate images; grid searched over [0.1, 0.25, 0.5] in Appendix C.
  • Meta-training learning rate = 0.001
    Adam learning rate for the meta-LoRA; grid searched over [0.001, 0.01, 0.1] in Appendix C.
  • Token sparsity ratio = 25%, 50%, or 75% depending on dataset
    The best sparsity per dataset is selected and reported with different suffixes (LoRA Recycle25/50/75 in Tables 2 and 3); the headline 6.27% gain mixes these choices.
  • LoRA rank r = 4
    Rank of source and meta LoRAs chosen by hand; a small ablation with mixed ranks shows robustness in Table 15.
assumptions (4)
  • domain assumption Pre-tuned LoRAs and new tasks are drawn from the same underlying task distribution pT, with disjoint label spaces but shared domain structure.
    Stated in Section 3 problem setup; this is a standard meta-learning assumption, but its validity drives the cross-domain results.
  • domain assumption Surrogate data generated by optimizing Eq. (1) preserve the class-discriminative structure of the original private training data.
    Core of LoRA Inversion in Section 4.1; tested indirectly via Table 9, but not guaranteed for complex real distributions.
  • ad hoc to paper BN statistics of an open-source ResNet50 can serve as a naturalness prior for CLIP ViT feature maps.
    Section 4.1: ViT has no BN layers, so statistics are borrowed from ResNet50 as suggested by ref. [20]; no validation is given that the two feature spaces align.
  • domain assumption Attention weights of the [CLS] token at the final layer identify the most informative tokens, so pruning them during inversion and masking them during meta-training preserves accuracy.
    Section 4.3; supported by ablations in Tables 2, 4, 5, and 11, but relies on an empirical regularity of self-attention.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unlocking Tuning-Free Few-Shot Adaptability in Visual Foundation Models by Recycling Pre-Tuned LoRAs." pith.science (2026). https://pith.science/paper/TBD5KIMF

@misc{pith2026241202220,
  author       = {Pith},
  title        = {Pith review of: Unlocking Tuning-Free Few-Shot Adaptability in Visual Foundation Models by Recycling Pre-Tuned LoRAs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TBD5KIMF}},
  note         = {Machine review of arXiv:2412.02220}
}
read the original abstract

Large Language Models (LLMs) such as ChatGPT demonstrate strong few-shot adaptability without requiring fine-tuning, positioning them ideal for data-limited and real-time applications. However, this adaptability has not yet been replicated in current Visual Foundation Models (VFMs), which require explicit fine-tuning with sufficient tuning data. Besides, the pretraining-finetuning paradigm has led to the surge of numerous task-specific modular components, such as Low-Rank Adaptation (LoRA). For the first time, we explore the potential of reusing diverse pre-tuned LoRAs without accessing their original training data, to achieve tuning-free few-shot adaptation in VFMs. Our framework, LoRA Recycle, distills a meta-LoRA from diverse pre-tuned LoRAs with a meta-learning objective, using surrogate data generated inversely from pre-tuned LoRAs themselves. The VFM, once equipped with the meta-LoRA, is empowered to solve new few-shot tasks in a single forward pass, akin to the in-context learning of LLMs. Additionally, we incorporate a double-efficient mechanism tailored to our framework, significantly accelerating the meta-training process while maintaining or even improving performance. Extensive experiments across various few-shot classification benchmarks across both in- and cross-domain scenarios demonstrate the superiority of our framework.

Figures

Figures reproduced from arXiv: 2412.02220 by the authors.

Figure 1
Figure 1. Thanks to the modularity of LoRA, users can upload [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of LoRA Recycle. (i) (Pink Path) We generate task-specific surrogate data from the pre-tuned LoRA via LoRA Inversion. The input data (attached with the fire in the left corner) is initialized as Gaussian noise and iteratively optimized by minimizing Ldata (Eq. (1)). The surrogate data is then used to construct a meta-training task with one support set and one query set. (ii) (Black Path) We meta-train the m… view at source ↗
Figure 3
Figure 3. Double-Efficient Mechanism. (Left: Inversion Stage) During the inversion stage, token pruning is applied in the hidden layers by removing unimportant tokens based on self-attention weights, accelerating both forward and backward computations for data generation. (Right: Meta-Training Stage) To highlight the most informative areas in the generated image, we construct a mask by setting values of 1 at the positions of … view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visualization of generated images with their 75% token-masked versions. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Visualization of generated images with (left) and without (right) the naturalness prior [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Visualization of generated images (odd line) and their 75% token-masked versions (even line) from ViT-B/32. (32 [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Visualization of masked generated images at varying sparsity levels. (32 [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: Comparison with SOTA model inversion approach. Our model inversion approach surpasses the current SOTA method CMI [ [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: T-SNE visualization of generated images. Our model inversion approach successfully inverts the essential discriminative features, [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 54 canonical work pages

  1. [1]

    Meta-adapters: Parameter ef- ficient few-shot fine-tuning through meta-learning

    Trapit Bansal, Salaheddin Alzubi, Tong Wang, Jay-Yoon Lee, and Andrew McCallum. Meta-adapters: Parameter ef- ficient few-shot fine-tuning through meta-learning. In Inter- national Conference on Automated Machine Learning, pages 19–1. PMLR, 2022. 2, 18

    work page 2022

  2. [2]

    Meta-learning with differentiable closed-form solvers

    Luca Bertinetto, Joao F Henriques, Philip HS Torr, and An- drea Vedaldi. Meta-learning with differentiable closed-form solvers. arXiv preprint arXiv:1805.08136, 2018. 6

    arXiv 2018

  3. [3]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Sub- biah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakan- tan, Pranav Shyam, Girish Sastry, Amanda Askell, Sand- hini Agarwal, Ariel Herbert-V oss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, S...

    1901

  4. [4]

    Cross-Domain Few-Shot Learning with Meta Fine-Tuning

    John Cai and Sheng Mei Shen. Cross-domain few-shot learn- ing with meta fine-tuning. arXiv preprint arXiv:2005.10544,

    work page Pith review arXiv 2005

  5. [5]

    Llava-mole: Sparse mixture of lora experts for mitigating data con- flicts in instruction finetuning mllms

    Shaoxiang Chen, Zequn Jie, and Lin Ma. Llava-mole: Sparse mixture of lora experts for mitigating data con- flicts in instruction finetuning mllms. arXiv preprint arXiv:2401.16160, 2024. 2, 6, 16

    arXiv 2024

  6. [6]

    Meta-learning via language model in-context tuning

    Yanda Chen, Ruiqi Zhong, Sheng Zha, George Karypis, and He He. Meta-learning via language model in-context tuning. In Proceedings of the 60th Annual Meeting of the Associa- tion for Computational Linguistics (Volume 1: Long Papers), pages 719–730, Dublin, Ireland, 2022. Association for Com- putational Linguistics. 18

    work page 2022

  7. [7]

    Scaling instruction-finetuned language models

    Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416, 2022. 18

    arXiv

  8. [8]

    Why can gpt learn in-context? language models implicitly perform gradient descent as meta-optimizers

    Damai Dai, Yutao Sun, Li Dong, Yaru Hao, Shuming Ma, Zhifang Sui, and Furu Wei. Why can gpt learn in-context? language models implicitly perform gradient descent as meta-optimizers. arXiv preprint arXiv:2212.10559, 2022. 3

    arXiv 2022

Show all 75 references
  1. [9]

    A survey on in-context learning

    Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Zhiyong Wu, Baobao Chang, Xu Sun, Jingjing Xu, and Zhifang Sui. A survey on in-context learning. arXiv preprint arXiv:2301.00234, 2022. 3

    2022 arXiv

  2. [10]

    Contrastive model inver- sion for data-free knowledge distillation

    Gongfan Fang, Jie Song, Xinchao Wang, Chengchao Shen, Xingen Wang, and Mingli Song. Contrastive model inver- sion for data-free knowledge distillation. arXiv preprint arXiv:2105.08584, 2021. 15, 17, 18

    2021 arXiv

  3. [11]

    Up to 100x faster data- free knowledge distillation

    Gongfan Fang, Kanya Mo, Xinchao Wang, Jie Song, Shitao Bei, Haofei Zhang, and Mingli Song. Up to 100x faster data- free knowledge distillation. InProceedings of the AAAI Con- ference on Artificial Intelligence, pages 6597–6604, 2022. 18

    2022

  4. [12]

    Context-aware meta-learning

    Christopher Fifty, Dennis Duan, Ronald G Junkins, Ehsan Amid, Jure Leskovec, Christopher R ´e, and Sebastian Thrun. Context-aware meta-learning. arXiv preprint arXiv:2310.10971, 2023. 3, 6, 17

    2023 arXiv

  5. [13]

    Model- agnostic meta-learning for fast adaptation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model- agnostic meta-learning for fast adaptation of deep networks. In International conference on machine learning , pages 1126–1135. PMLR, 2017. 2, 3, 4, 6, 18

    2017

  6. [14]

    Styleadv: Meta style adversarial training for cross-domain few-shot learning

    Yuqian Fu, Yu Xie, Yanwei Fu, and Yu-Gang Jiang. Styleadv: Meta style adversarial training for cross-domain few-shot learning. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 24575–24584, 2023. 2

    2023

  7. [15]

    On the effectiveness of parameter-efficient fine-tuning

    Zihao Fu, Haoran Yang, Anthony Man-Cho So, Wai Lam, Lidong Bing, and Nigel Collier. On the effectiveness of parameter-efficient fine-tuning. In Proceedings of the AAAI Conference on Artificial Intelligence , pages 12799–12807,

  8. [16]

    Clip-adapter: Better vision-language models with feature adapters

    Peng Gao, Shijie Geng, Renrui Zhang, Teli Ma, Rongyao Fang, Yongfeng Zhang, Hongsheng Li, and Yu Qiao. Clip-adapter: Better vision-language models with feature adapters. International Journal of Computer Vision , pages 1–15, 2023. 2

    2023

  9. [17]

    Know where you’re going: Meta-learning for parameter-efficient fine-tuning

    Mozhdeh Gheini, Xuezhe Ma, and Jonathan May. Know where you’re going: Meta-learning for parameter-efficient fine-tuning. In Findings of the Association for Computa- tional Linguistics: ACL 2023, pages 11602–11612, Toronto, Canada, 2023. Association for Computational Linguistics. 18

    2023

  10. [18]

    Mixture of cluster-conditional lora experts for vision-language instruction tuning

    Yunhao Gou, Zhili Liu, Kai Chen, Lanqing Hong, Hang Xu, Aoxue Li, Dit-Yan Yeung, James T Kwok, and Yu Zhang. Mixture of cluster-conditional lora experts for vision-language instruction tuning. arXiv preprint arXiv:2312.12379, 2023. 2

    2023 arXiv

  11. [19]

    A broader study of cross-domain few-shot learning

    Yunhui Guo, Noel C Codella, Leonid Karlinsky, James V Codella, John R Smith, Kate Saenko, Tajana Rosing, and Rogerio Feris. A broader study of cross-domain few-shot learning. In European conference on computer vision, pages 124–141. Springer, 2020. 7

    2020

  12. [20]

    Gradvit: 9 Gradient inversion of vision transformers

    Ali Hatamizadeh, Hongxu Yin, Holger R Roth, Wenqi Li, Jan Kautz, Daguang Xu, and Pavlo Molchanov. Gradvit: 9 Gradient inversion of vision transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10021–10030, 2022. 4

    2022

  13. [21]

    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

    2021 arXiv

  14. [22]

    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. In International Con- ference on Learning Representations, 2022. 2

    2022

  15. [23]

    Revisiting data-free knowledge distilla- tion with poisoned teachers

    Junyuan Hong, Yi Zeng, Shuyang Yu, Lingjuan Lyu, Ruoxi Jia, and Jiayu Zhou. Revisiting data-free knowledge distilla- tion with poisoned teachers. In International Conference on Machine Learning, pages 13199–13212. PMLR, 2023. 18

    2023

  16. [24]

    Meta- learning the difference: preparing large language models for efficient adaptation

    Zejiang Hou, Julian Salazar, and George Polovets. Meta- learning the difference: preparing large language models for efficient adaptation. Transactions of the Association for Computational Linguistics, 10:1249–1265, 2022. 18

    2022

  17. [25]

    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. 2

    2019

  18. [26]

    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. 1, 2, 3, 6, 16

    2021 arXiv

  19. [27]

    Pushing the limits of simple pipelines for few-shot learning: External data and fine-tuning make a difference

    Shell Xu Hu, Da Li, Jan St ¨uhmer, Minyoung Kim, and Tim- othy M Hospedales. Pushing the limits of simple pipelines for few-shot learning: External data and fine-tuning make a difference. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pa...

    2022

  20. [28]

    Sparse model inversion: Ef- ficient inversion of vision transformers for data-free appli- cations

    Zixuan Hu, Yongxian Wei, Li Shen, Zhenyi Wang, Lei Li, Chun Yuan, and Dacheng Tao. Sparse model inversion: Ef- ficient inversion of vision transformers for data-free appli- cations. In Forty-first International Conference on Machine Learning. 18

  21. [29]

    Architecture, dataset and model- scale agnostic data-free meta-learning

    Zixuan Hu, Li Shen, Zhenyi Wang, Tongliang Liu, Chun Yuan, and Dacheng Tao. Architecture, dataset and model- scale agnostic data-free meta-learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023. 3, 18, 19

    2023

  22. [30]

    Learning to learn from apis: Black-box data-free meta-learning

    Zixuan Hu, Li Shen, Zhenyi Wang, Baoyuan Wu, Chun Yuan, and Dacheng Tao. Learning to learn from apis: Black-box data-free meta-learning. arXiv preprint arXiv:2305.18413, 2023. 3, 18, 19

    2023 arXiv

  23. [31]

    Lorahub: Efficient cross-task gen- eralization via dynamic lora composition

    Chengsong Huang, Qian Liu, Bill Yuchen Lin, Tianyu Pang, Chao Du, and Min Lin. Lorahub: Efficient cross-task gen- eralization via dynamic lora composition. arXiv preprint arXiv:2307.13269, 2023. 2, 6, 16

    2023 arXiv

  24. [32]

    Diversity-aware meta visual prompting

    Qidong Huang, Xiaoyi Dong, Dongdong Chen, Weim- ing Zhang, Feifei Wang, Gang Hua, and Nenghai Yu. Diversity-aware meta visual prompting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10878–10887, 2023. 18

    2023

  25. [33]

    Opt-iml: Scaling language model instruction meta learning through the lens of generalization

    Srinivasan Iyer, Xi Victoria Lin, Ramakanth Pasunuru, Todor Mihaylov, Daniel Simig, Ping Yu, Kurt Shuster, Tianlu Wang, Qing Liu, Punit Singh Koura, et al. Opt-iml: Scaling language model instruction meta learning through the lens of generalization. arXiv preprint arXiv:2212.1...

    2022 arXiv

  26. [34]

    Rethinking efficient tun- ing methods from a unified perspective

    Zeyinzi Jiang, Chaojie Mao, Ziyuan Huang, Yiliang Lv, Deli Zhao, and Jingren Zhou. Rethinking efficient tun- ing methods from a unified perspective. arXiv preprint arXiv:2303.00690, 2023. 2

    2023 arXiv

  27. [35]

    All tokens matter: Token labeling for training better vision transform- ers

    Zi-Hang Jiang, Qibin Hou, Li Yuan, Daquan Zhou, Yujun Shi, Xiaojie Jin, Anran Wang, and Jiashi Feng. All tokens matter: Token labeling for training better vision transform- ers. Advances in neural information processing systems, 34: 18590–18602, 2021. 14

    2021

  28. [36]

    Adaptive gradient-based meta-learning methods.Ad- vances in Neural Information Processing Systems, 32, 2019

    Mikhail Khodak, Maria-Florina F Balcan, and Ameet S Tal- walkar. Adaptive gradient-based meta-learning methods.Ad- vances in Neural Information Processing Systems, 32, 2019. 2

    2019

  29. [37]

    Token fusion: Bridging the gap between token pruning and token merging

    Minchul Kim, Shangqian Gao, Yen-Chang Hsu, Yilin Shen, and Hongxia Jin. Token fusion: Bridging the gap between token pruning and token merging. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 1383–1392, 2024. 13

    2024

  30. [38]

    Fine-tuning can distort pretrained fea- tures and underperform out-of-distribution

    Ananya Kumar, Aditi Raghunathan, Robbie Jones, Tengyu Ma, and Percy Liang. Fine-tuning can distort pretrained fea- tures and underperform out-of-distribution. arXiv preprint arXiv:2202.10054, 2022. 13

    2022 arXiv

  31. [39]

    Surgical fine- tuning improves adaptation to distribution shifts

    Yoonho Lee, Annie S Chen, Fahim Tajwar, Ananya Kumar, Huaxiu Yao, Percy Liang, and Chelsea Finn. Surgical fine- tuning improves adaptation to distribution shifts. Interna- tional Conference on Learning Representations, 2023. 18

    2023

  32. [40]

    Patch similarity aware data-free quantization for vision transformers

    Zhikai Li, Liping Ma, Mengjuan Chen, Junrui Xiao, and Qingyi Gu. Patch similarity aware data-free quantization for vision transformers. In European Conference on Computer Vision, pages 154–170, 2022. 18

    2022

  33. [41]

    Psaq-vit v2: Toward accurate and general data-free quanti- zation for vision transformers

    Zhikai Li, Mengjuan Chen, Junrui Xiao, and Qingyi Gu. Psaq-vit v2: Toward accurate and general data-free quanti- zation for vision transformers. IEEE Transactions on Neural Networks and Learning Systems, 2023. 18

    2023

  34. [42]

    Training-free open-ended object detection and segmentation via attention as prompts

    Zhiwei Lin, Yongtao Wang, and Zhi Tang. Training-free open-ended object detection and segmentation via attention as prompts. arXiv preprint arXiv:2410.05963, 2024

    2024 arXiv

  35. [43]

    Few- shot parameter-efficient fine-tuning is better and cheaper than in-context learning

    Haokun Liu, Derek Tam, Mohammed Muqeeth, Jay Mohta, Tenghao Huang, Mohit Bansal, and Colin A Raffel. Few- shot parameter-efficient fine-tuning is better and cheaper than in-context learning. Advances in Neural Information Processing Systems, 35:1950–1965, 2022. 2

    1950

  36. [44]

    Small scale data-free knowledge distillation

    He Liu, Yikai Wang, Huaping Liu, Fuchun Sun, and Anbang Yao. Small scale data-free knowledge distillation. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6008–6016, 2024. 18

    2024

  37. [45]

    Matcher: Segment anything with one shot using all-purpose feature matching

    Yang Liu, Muzhi Zhu, Hengtao Li, Hao Chen, Xinlong Wang, and Chunhua Shen. Matcher: Segment anything with one shot using all-purpose feature matching. In The Twelfth International Conference on Learning Representa- tions, 2024. 3

    2024

  38. [46]

    DFRD: Data-free robustness distillation 10 for heterogeneous federated learning

    Kangyang Luo, Shuai Wang, Yexuan Fu, Xiang Li, Yunshi Lan, and Ming Gao. DFRD: Data-free robustness distillation 10 for heterogeneous federated learning. In Thirty-seventh Con- ference on Neural Information Processing Systems, 2023. 18

    2023

  39. [47]

    MetaICL: Learning to learn in context

    Sewon Min, Mike Lewis, Luke Zettlemoyer, and Hannaneh Hajishirzi. MetaICL: Learning to learn in context. In Pro- ceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2791–2809, Seattl...

    2022

  40. [48]

    Meta learning to bridge vision and language models for mul- timodal few-shot learning

    Ivona Najdenkoska, Xiantong Zhen, and Marcel Worring. Meta learning to bridge vision and language models for mul- timodal few-shot learning. In The Eleventh International Conference on Learning Representations, 2023. 18

    2023

  41. [49]

    On first-order meta-learning algorithms

    Alex Nichol, Joshua Achiam, and John Schulman. On first-order meta-learning algorithms. arXiv preprint arXiv:1803.02999, 2018. 4, 18

    2018 arXiv

  42. [50]

    Automated flower classification over a large number of classes

    Maria-Elena Nilsback and Andrew Zisserman. Automated flower classification over a large number of classes. In 2008 Sixth Indian conference on computer vision, graphics & im- age processing, pages 722–729. IEEE, 2008. 6

    2008

  43. [51]

    Dynamicvit: Efficient vision transformers with dynamic token sparsification

    Yongming Rao, Wenliang Zhao, Benlin Liu, Jiwen Lu, Jie Zhou, and Cho-Jui Hsieh. Dynamicvit: Efficient vision transformers with dynamic token sparsification. Advances in neural information processing systems, 34:13937–13949,

  44. [52]

    Data-free knowledge distillation for fine-grained visual cat- egorization

    Renrong Shao, Wei Zhang, Jianhua Yin, and Jun Wang. Data-free knowledge distillation for fine-grained visual cat- egorization. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 1515–1525, 2023. 18

    2023

  45. [53]

    Prototypical networks for few-shot learning

    Jake Snell, Kevin Swersky, and Richard Zemel. Prototypical networks for few-shot learning. Advances in neural informa- tion processing systems, 30, 2017. 15

    2017

  46. [54]

    Meta-transfer learning for few-shot learning

    Qianru Sun, Yaoyao Liu, Tat-Seng Chua, and Bernt Schiele. Meta-transfer learning for few-shot learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 403–412, 2019. 18

    2019

  47. [55]

    Training data-efficient image transformers & distillation through at- tention

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herv ´e J´egou. Training data-efficient image transformers & distillation through at- tention. In International conference on machine learning , pages 10347–10357. PMLR, 2021. 14

    2021

  48. [56]

    Large-scale data-free knowledge distillation for imagenet via multi-resolution data generation

    Minh-Tuan Tran, Trung Le, Xuan-May Le, Jianfei Cai, Mehrtash Harandi, and Dinh Phung. Large-scale data-free knowledge distillation for imagenet via multi-resolution data generation. arXiv preprint arXiv:2411.17046, 2024. 18

    2024 arXiv

  49. [57]

    Nayer: Noisy layer data generation for efficient and effective data-free knowl- edge distillation

    Minh-Tuan Tran, Trung Le, Xuan-May Le, Mehrtash Ha- randi, Quan Hung Tran, and Dinh Phung. Nayer: Noisy layer data generation for efficient and effective data-free knowl- edge distillation. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition ...

    2024

  50. [58]

    Meta-dataset: A dataset of datasets for learning to learn from few examples

    Eleni Triantafillou, Tyler Zhu, Vincent Dumoulin, Pascal Lamblin, Utku Evci, Kelvin Xu, Ross Goroshin, Carles Gelada, Kevin Jordan Swersky, Pierre-Antoine Manzagol, and Hugo Larochelle. Meta-dataset: A dataset of datasets for learning to learn from few examples. In Internation...

    2020

  51. [59]

    Matching networks for one shot learning.Ad- vances in neural information processing systems , 29, 2016

    Oriol Vinyals, Charles Blundell, Timothy Lillicrap, Daan Wierstra, et al. Matching networks for one shot learning.Ad- vances in neural information processing systems , 29, 2016. 6

    2016

  52. [60]

    Transformers learn in-context by gradient descent

    Johannes V on Oswald, Eyvind Niklasson, Ettore Randazzo, Jo˜ao Sacramento, Alexander Mordvintsev, Andrey Zhmogi- nov, and Max Vladymyrov. Transformers learn in-context by gradient descent. In International Conference on Machine Learning, pages 35151–35174. PMLR, 2023. 3

    2023

  53. [61]

    The Caltech-UCSD Birds-200- 2011 Dataset

    Catherine Wah, Steve Branson, Peter Welinder, Pietro Per- ona, and Serge Belongie. The Caltech-UCSD Birds-200- 2011 Dataset. 2011. 6

    2011

  54. [62]

    Generalizing to unseen domains: A survey on do- main generalization

    Jindong Wang, Cuiling Lan, Chang Liu, Yidong Ouyang, Tao Qin, Wang Lu, Yiqiang Chen, Wenjun Zeng, and S Yu Philip. Generalizing to unseen domains: A survey on do- main generalization. IEEE transactions on knowledge and data engineering, 35(8):8052–8072, 2022. 18

    2022

  55. [63]

    De-confounded data-free knowledge distillation for handling distribution shifts

    Yuzheng Wang, Dingkang Yang, Zhaoyu Chen, Yang Liu, Siao Liu, Wenqiang Zhang, Lihua Zhang, and Lizhe Qi. De-confounded data-free knowledge distillation for handling distribution shifts. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pag...

    2024

  56. [64]

    Meta learning on a sequence of imbalanced domains with difficulty awareness

    Zhenyi Wang, Tiehang Duan, Le Fang, Qiuling Suo, and Mingchen Gao. Meta learning on a sequence of imbalanced domains with difficulty awareness. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 8947–8957, 2021. 3

    2021

  57. [65]

    Meta-learning without data via wasserstein distributionally- robust model fusion

    Zhenyi Wang, Xiaoyang Wang, Li Shen, Qiuling Suo, Kaiqiang Song, Dong Yu, Yan Shen, and Mingchen Gao. Meta-learning without data via wasserstein distributionally- robust model fusion. In The 38th Conference on Uncertainty in Artificial Intelligence, 2022. 18

    2022

  58. [66]

    Task groupings regular- ization: Data-free meta-learning with heterogeneous pre- trained models

    Yongxian Wei, Zixuan Hu, Li Shen, Zhenyi Wang, Yu Li, Chun Yuan, and Dacheng Tao. Task groupings regular- ization: Data-free meta-learning with heterogeneous pre- trained models. In Forty-first International Conference on Machine Learning, 2024. 3

    2024

  59. [67]

    Free: Faster and better data-free meta-learning

    Yongxian Wei, Zixuan Hu, Zhenyi Wang, Li Shen, Chun Yuan, and Dacheng Tao. Free: Faster and better data-free meta-learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 23273– 23282, 2024. 3

    2024

  60. [68]

    pi-tuning: Transferring multimodal foundation models with optimal multi-task inter- polation

    Chengyue Wu, Teng Wang, Yixiao Ge, Zeyu Lu, Ruisong Zhou, Ying Shan, and Ping Luo. pi-tuning: Transferring multimodal foundation models with optimal multi-task inter- polation. In International Conference on Machine Learning, pages 37713–37727. PMLR, 2023. 2

    2023

  61. [69]

    Mole: Mixture of lora experts

    Xun Wu, Shaohan Huang, and Furu Wei. Mole: Mixture of lora experts. In The Twelfth International Conference on Learning Representations, 2023. 2

    2023

  62. [70]

    Meta-personalizing vision- language models to find named instances in video

    Chun-Hsiao Yeh, Bryan Russell, Josef Sivic, Fabian Caba Heilbron, and Simon Jenni. Meta-personalizing vision- language models to find named instances in video. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19123–19132, 2023. 18

    2023

  63. [71]

    11 Dreaming to distill: Data-free knowledge transfer via deep- inversion

    Hongxu Yin, Pavlo Molchanov, Jose M Alvarez, Zhizhong Li, Arun Mallya, Derek Hoiem, Niraj K Jha, and Jan Kautz. 11 Dreaming to distill: Data-free knowledge transfer via deep- inversion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , page...

    2020

  64. [72]

    Bayesian model-agnostic meta-learning

    Jaesik Yoon, Taesup Kim, Ousmane Dia, Sungwoong Kim, Yoshua Bengio, and Sungjin Ahn. Bayesian model-agnostic meta-learning. Advances in neural information processing systems, 31, 2018. 2

    2018

  65. [73]

    Data-free knowledge distillation via feature exchange and activation region constraint

    Shikang Yu, Jiachen Chen, Hu Han, and Shuqiang Jiang. Data-free knowledge distillation via feature exchange and activation region constraint. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 24266–24275, 2023. 18

    2023

  66. [74]

    recycle in-domain LoRAs

    Renrui Zhang, Zhengkai Jiang, Ziyu Guo, Shilin Yan, Junt- ing Pan, Hao Dong, Yu Qiao, Peng Gao, and Hongsheng Li. Personalize segment anything model with one shot. In The Twelfth International Conference on Learning Representa- tions, 2024. 3 12 Appendix A. Additional Experime...

    2024

  67. [75]

    and model quantization [40, 41]. Unlike DFKD, which primarily employs inverted data to distill knowledge from a single teacher model, our study in- troduces a meta-learning framework that harnesses inverted data across multiple teacher models. Moreover, instead of transferring...

Pith tools

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