Latest Advancements Towards Catastrophic Forgetting under Data Scarcity: A Comprehensive Survey on Few-Shot Class Incremental Learning

Igor Skrjanc; Mahardhika Pratama; M. Anwar Ma'sum

REVIEW 5 major objections 5 minor 54 references

Latest Advancements Towards Catastrophic Forgetting under Data Scarcity: A Comprehensive Survey on Few-Shot Class Incremental Learning

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

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

Pith's one-line read This survey structures few-shot class-incremental learning into five approaches and five data settings, and argues that prompt-based parameter-efficient tuning on pre-trained models now leads the field.

desk verdict A genuinely useful organizing topology for FSCIL, but the PEFT-superiority headline is confounded and the survey needs methodological tightening. read the letter →

arxiv 2502.08181 v1 pith:6TWOZQKF submitted 2025-02-12 cs.LG cs.AIcs.CV

M. Anwar Ma'sum , Mahardhika Pratama , Igor Skrjanc This is my paper · ORCID

classification cs.LGcs.AIcs.CV

keywords few-shotclass-incrementallearningcatastrophicforgettingprototyperectificationparameter-efficientfine-tuningpromptpre-trainedvision-languagemodelcontinualsurveystability-plasticity

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to organize the sprawling few-shot class-incremental learning (FSCIL) literature into a clear topology: five method families (backbone tuning, meta-learning, prototype tuning, dynamic architecture, and parameter-efficient fine-tuning) and five problem settings (supervised, semi-supervised, unsupervised, complete, and cross-domain). It gives each family a formal objective and compares methods not only on average accuracy but also on stability-plasticity balance and base-versus-novel class performance. The survey's main empirical claim is that PEFT methods built on frozen pre-trained vision transformers outperform the other four families by a margin of up to 20% average accuracy, while using far fewer trainable parameters. It also argues that prototype rectification is a central and under-appreciated mechanism for countering data scarcity. A reliable map of this kind matters because it tells researchers which directions are paying off and where the open problems actually lie.

What carries the argument

The central organizing device is a five-way taxonomy of FSCIL approaches, each with an explicit objective function (equations 2 through 7), paired with a classification of problem settings (fully supervised, semi-supervised, unsupervised, complete, and cross-domain). Within the taxonomy, the load-bearing concept is the prototype-based classifier, which stores one vector per class and compares new inputs against those prototypes; unlike network classifiers, prototypes are not updated in later tasks, so they are naturally resistant to forgetting. The paper's second machinery is a topology of prototype rectification, distinguishing loss-driven updates, pseudo-prototype generation, training-free calibration, and trainable calibrator networks, and it uses this taxonomy to explain why PEFT-plus-prototype hybrids such as L2P+, DualP+, and CodaP+ markedly outperform their original prompt-only versions.

What would settle it

A controlled re-benchmarking study that fixes the backbone (e.g., ViT-B/16 for all methods), fixes the pre-trained model, and uses the same session splits and evaluation code for at least one representative method from each of the five families would settle whether the up-to-20% PEFT advantage holds; if the gap shrinks to near zero, the survey's main ranking claim collapses.

Watch

Extended reading notes

Core claim

The paper argues that the FSCIL field has moved through four earlier approaches that tune or extend a backbone trained on a large base task, and into a fifth, PEFT-based approach that keeps a pre-trained transformer frozen and learns only small prompt parameters. It finds that PEFT methods not only reduce trainable parameters from tens of millions to under a million but also lift average accuracy substantially, with the best results exceeding 90% on MiniImageNet and 80% on CUB, because the pre-trained model supplies the general knowledge that the few-shot tasks cannot provide. At the same time, the survey shows that almost all methods, including the best PEFT ones, still achieve far lower harmonic mean accuracy than average accuracy, indicating that plasticity on novel classes remains weak and that stability-plasticity is still an open problem. It further identifies prototype rectification—correcting biased class prototypes estimated from very few samples—as a key mechanism, and classifies existing rectification strategies into four types: loss-based, pseudo-prototype, training-free calibration, and trainable-network calibration.

Load-bearing premise

The survey's central comparisons assume that the accuracy numbers collected from different papers are directly comparable even though the methods use different backbones, different pre-trained models or none, and possibly different evaluation protocols.

Editorial extensions

If this is right

If the PEFT advantage is real, future FSCIL work should default to frozen pre-trained backbones with prompt learning and a prototype-based classifier, rather than tuning full backbones from scratch.
Prototype rectification becomes a priority design choice, with training-free calibration offering a low-cost baseline that many current PEFT methods do not yet exploit.
Reporting average accuracy alone is misleading; the survey's stability-plasticity analysis implies that base-versus-novel class accuracy and harmonic mean accuracy should become standard reporting metrics.
The five-setting taxonomy gives practitioners a way to match a method to their actual data regime, e.g., choosing supervised FSCIL when labels are complete and cross-domain or semi-supervised variants when they are not.
Open challenges identified in the survey—federated FSCIL, online data streams, open-world detection, and class imbalance—point to concrete next problems that the current methods are not yet built to handle.

Reading between the lines

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

The claimed up-to-20% gap between PEFT and the other approaches is likely confounded by base model scale and pre-training data: PEFT methods use ViT-B/16, which was itself pre-trained on ImageNet, the superset of the MiniImageNet test set, so part of the gain may be transfer leakage rather than the prompting mechanism per se.
A controlled comparison that keeps backbone, pre-training, and protocol identical across the five families would be the natural next experiment, and the survey's own table does not currently enable that because all non-PEFT methods use ResNet18.
The observation that almost no applied FSCIL system uses PEFT or language guidance suggests that the technique gap in real-world domains such as medical imaging and audio may be larger than the benchmark gap, and that transferring the PEFT-plus-prototype recipe to those domains is a testable extension.
The four-way rectification topology could be reused as an evaluation axis for any few-shot or continual-learning method, not only FSCIL, since biased prototypes are a general consequence of data scarcity.

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.

Desk Editor's Note

read the letter

Roughly put: the survey is a useful map, and the taxonomy is worth having, but the headline performance claim about PEFT superiority is not supported by the table it is based on. PEFT methods run on ImageNet-pretrained ViT-B/16 while the other four approaches mostly use ResNet18 trained from scratch (except on CUB), so the up-to-20% gap is almost certainly confounded with backbone and pretraining. The paper even says this about the MiniImageNet numbers, then draws the contrary conclusion.

What is genuinely good: the five-approach taxonomy (backbone tuning, meta-learning, prototype tuning, dynamic architecture, PEFT) and the five-settings classification (supervised, semi-supervised, unsupervised, complete, cross-domain) organize the literature better than the earlier Tian et al. survey. The prototype-rectification categorization into loss, pseudo-prototype, training-free calibration, and trainable network is a nice lens. The coverage of recent PEFT and language-guided methods (ASP, PriViLege, FineFMPL, ApproxFSCIL) is timely, and the attention to base vs novel class accuracy and AHM is a real service to the community.

Soft spots, in order of weight. First, the PEFT superiority claim in Section 4.6 is load-bearing and confounded. The table includes L2P, DualP, and CodaP—rehearsal-free CIL methods, not FSCIL methods—and Table 2 shows 0% novel-class accuracy for them. That weakens the comparison further. Second, the survey has no methodology section; there are no inclusion/exclusion criteria, so 'comprehensive' is unverifiable. Third, the claimed coverage of SSFSCIL is not backed: it is defined in Section 2 but no SSFSCIL method appears in Tables 1 or 3. Fourth, there are small internal errors: a cross-reference to Eq. 2 when Eq. 4 is meant, a sentence cut off at the end of Section 4.7, and some garbled phrases. Table 1 has blanks and no variance/significance information, which the authors should acknowledge.

The formal objectives (Eqs. 2-5) are near-restatements of Eq. 1 with different parameter sets; that is fine as a pedagogical device but not a mathematical contribution. The self-citations to the authors' own UFSCIL and FFSCIL work are not a problem per se, but they do shape the 'future directions' section; a reader should know those are not established subfields yet.

Who is this for? Someone new to FSCIL who wants a structured map and a recent pointer list. A serious referee should engage with it, mostly to force the performance claims to be re-analyzed and the missing methodology made explicit. I would not desk-reject, but I would not accept as is.

Referee Report

5 major / 5 minor

Summary. This manuscript is a survey of few-shot class incremental learning (FSCIL). It proposes a five-way taxonomy of approaches (backbone tuning, meta-learning, prototype tuning, dynamic architecture, and parameter-efficient fine-tuning), formal objectives for each approach, and a taxonomy of five settings (supervised, semi-supervised, unsupervised, complete, and cross-domain FSCIL). The paper compiles a large comparison table (Table 1) covering backbone, learnable parameters, pre-trained model usage, classifier type, augmentation, prototype rectification, language guidance, and accuracy metrics (AA, PD, AHM), and it includes a separate application table (Table 3) for medical, remote sensing, NLP, and graph domains. It also discusses prototype rectification mechanisms, stability-plasticity trade-offs, open challenges, and future directions. The paper's central practical claim is that PEFT methods outperform the other four approaches by a significant margin, up to 20%.

Significance. If the comparison were methodologically matched, the survey would be a useful map of the FSCIL field: Table 1 is unusually detailed, the formal objective statements for each approach are a useful reference, and the application table covers areas that are often omitted from FSCIL surveys. The paper does not provide machine-checked proofs or reproducible code, but its contribution is descriptive and organizational. However, the headline trend is currently confounded: the PEFT rows use a different backbone and pretraining regime from the other approaches, so the reported accuracy gap cannot be attributed to the PEFT approach alone. With a corrected comparison and more cautious claims, the survey could serve as a valuable reference; in its present form, its main conclusion is not supported by the evidence it presents.

major comments (5)

[§4.6, Table 1] The claim that 'PEFT methods e.g. ASP and Privilege confirm the superiority of the PEFT approach to the other 4 approaches with a significant gap i.e. up to 20%' is confounded. In Table 1, all PEFT rows use ViT-B/16 (86.57M parameters) with a pre-trained model on all datasets, whereas the other four approaches use ResNet18 (11.67M parameters) and a PTM only on CUB. The paper itself states in §4.6 that the >90% MiniImageNet results are 'common-sense results as the methods utilize pre-trained ViT on ImageNet (the superset of MiniImageNet)'. Therefore the accuracy gap may reflect backbone capacity and pretraining rather than the PEFT approach itself. The authors should provide a matched comparison with the same backbone, pretraining, sessions, and protocol, or substantially soften the causal wording of the PEFT-superiority conclusion.
[§3.3, Eq. (4)] The text says 'The objective of the prototype tuning approach is defined in eq. 2,' but Eq. (2) is the backbone-tuning objective; the prototype-tuning objective is Eq. (4). This is not merely a typo: the formal objective is one of the paper's stated contributions, and the incorrect pointer makes the section difficult to verify.
[§4.7, Table 2] Section 4.7 ends mid-sentence ('indicating that the') before Table 2, and the resumed text ('methods rely too much on base class accuracy') appears only after the table. The stability-plasticity analysis is a central part of the survey, and this incomplete sentence disrupts the presentation. The passage should be rewritten as complete prose so that Table 2 is cited rather than inserted into a sentence.
[§2.2, §4.8] The taxonomy in §2.2 lists semi-supervised FSCIL (SSFSCIL) as one of the sub-settings, and §1 claims coverage of SSFSCIL, but no surveyed method in Table 1 or Table 3 actually uses the semi-supervised setting. In §4.8, the only non-supervised settings mentioned are UFSCIL, CFSCIL, and CDFSCIL. The authors should either include SSFSCIL methods in the survey or explicitly state that the setting is defined for completeness but currently has no surveyed implementations.
[§3.5, Table 1, Table 2] L2P, DualP, and CodaP are presented as PEFT FSCIL methods, but Table 2 reports 0.00% novel-class accuracy for them on both CIFAR100 and CUB, which suggests they are not performing FSCIL as defined in §2.1. Their inclusion in the PEFT comparison, and their role as the base for the L2P+, DualP+, and CodaP+ variants, needs explicit justification. As presented, these rows weaken the comparison and the conclusion that PEFT is superior for FSCIL.

minor comments (5)

[§2.2 heading] The heading 'Types of FSCIL Setings' contains a typo; it should be 'Settings'.
[§3.1] The word 'aporoaches' should be 'approaches', and 'discrimivative' elsewhere should be 'discriminative'.
[§1] The term 'SSFCIL' is used in the introduction but 'SSFSCIL' is used elsewhere; the abbreviation should be consistent.
[§2.2(d), Table 1] The label 'CFSCIL' is used both for the 'Complete FSCIL' setting in §2.2(d) and for the method 'C-FSCIL' in Table 1. This ambiguity should be resolved, for example by renaming the setting or explicitly distinguishing the method name from the setting name.
[Table 1] The table reports accuracy numbers from different papers without variance or statistical significance information, and the 'Learn. Param.' column contains dashes for several PEFT methods. At minimum, the caption should state that the values are copied from the original papers and are not directly comparable; ideally, the authors would add citations or footnotes for each row.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the survey's claims are taxonomical and empirical, and the few author self-citations are not load-bearing.

full rationale

This is a survey paper with no derivation chain: it offers a taxonomy of FSCIL approaches, formal objectives, a comparison table, and qualitative findings. None of its central claims is defined in terms of a quantity fitted within the paper. The formal objectives in Eqs. (1)-(7) restate published method families and do not reduce to each other by construction. The paper's headline empirical finding, that PEFT methods show up to a 20% accuracy gap over the other four approaches (Section 4.6), is a transcription and comparison of published accuracy numbers. Even if that comparison is confounded by backbone and pretraining, as the paper itself acknowledges when it calls the >90% MiniImageNet results 'common-sense results as the methods utilize pre-trained ViT on ImageNet,' that is a validity or fairness concern, not a circularity: the claim is not equivalent to its inputs by definition. The only self-citations are to the authors' own earlier work on UFSCIL [Ma'Sum et al., 2024b], FLOWER [Ma'sum et al., 2024a], FFSCIL [Ma'sum et al., 2025a], and language-guided continual learning [Ma'sum et al., 2025b]. These are used to name sub-settings or to point to future directions, not to justify the accuracy comparisons or the proposed topology. The survey also explicitly flags setting fairness as an open problem in Section 6(f), showing that the comparability limitation is stated rather than hidden. Thus no load-bearing step reduces to a self-citation or to a fitted input, and the paper is self-contained as a survey relative to external published methods.

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

The survey introduces no new physical or mathematical entities; its inventions are taxonomical labels and problem-setting names, which are not independent entities. The free-parameter list is empty because the paper performs no fitting or derivations.

assumptions (3)

domain assumption FSCIL methods can be cleanly partitioned into five approaches (backbone tuning, meta-learning, prototype tuning, dynamic architecture, PEFT) and five settings (supervised, semi-supervised, unsupervised, complete, cross-domain).
Sections 2.2 and 3 impose this taxonomy; it is a modeling choice, not derived from the literature, and may misclassify hybrid methods.
domain assumption Reported accuracy numbers in Table 1 are faithful transcriptions from the cited papers and are comparable across methods.
Table 1 and Section 4.6 rely on these numbers for trend claims; no errors, variances, or protocol differences are shown.
domain assumption The 0/1 classification loss in Eq. 1 is an appropriate formalization of the FSCIL objective.
The survey uses this loss to define all approach objectives; other losses and metrics are not incorporated into the formal definitions.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Latest Advancements Towards Catastrophic Forgetting under Data Scarcity: A Comprehensive Survey on Few-Shot Class Incremental Learning." pith.science (2026). https://pith.science/paper/6TWOZQKF

@misc{pith2026250208181,
  author       = {Pith},
  title        = {Pith review of: Latest Advancements Towards Catastrophic Forgetting under Data Scarcity: A Comprehensive Survey on Few-Shot Class Incremental Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6TWOZQKF}},
  note         = {Machine review of arXiv:2502.08181}
}

read the original abstract

Data scarcity significantly complicates the continual learning problem, i.e., how a deep neural network learns in dynamic environments with very few samples. However, the latest progress of few-shot class incremental learning (FSCIL) methods and related studies show insightful knowledge on how to tackle the problem. This paper presents a comprehensive survey on FSCIL that highlights several important aspects i.e. comprehensive and formal objectives of FSCIL approaches, the importance of prototype rectifications, the new learning paradigms based on pre-trained model and language-guided mechanism, the deeper analysis of FSCIL performance metrics and evaluation, and the practical contexts of FSCIL in various areas. Our extensive discussion presents the open challenges, potential solutions, and future directions of FSCIL.

Figures

Figures reproduced from arXiv: 2502.08181 by the authors.

**Figure 1.** Types of FSCIL settings and their properties (a) Fully-Supervised, (b) Semi-Supervised (c) Unsupervised (d) Complete (e) CrossDomain FSCIL • We propose a comprehensive FSCIL topology highlighting the latest advancement of FSCIL methods, especially the new PEFT approach, the formal objective of each approach, and sub-settings of FSCIL i.e. SSFSCIL, UFSCIL, CFSCIL, and CDFSCIL. • We analyze broader aspects of FSCIL… view at source ↗

**Figure 2.** Visualization of FSCIL Approaches i.e. (a) backbone tuning approach, (b) meta learning approach, (c) prototype tuning approach, (d) dynamic structure approach, and (e) parameter efficient fine tuning (PEFT) approach. samples. Similarly, LDC[Liu et al., 2023] enforces a learnable models to approximate the distribution. F2M[Shi et al., 2021] and FLOWER[Ma’sum et al., 2024a] create a flat loss region by perturbating l… view at source ↗

**Figure 3.** Biased Prototype, Optimal Prototypes, and Types of Prototype Rectification Mechanism (a) Biased and Optimal Prototype (b) Rectification by Loss function only (c) by pseudo-prototypes (d) by training-free calibration (e) by trainable networks validated via a visual plot, while feature-space augmentation should be evaluated by valid metrics such as Euclidean distance, or loss. GP-tree augments its likelihood estimato… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 50 canonical work pages

[1]

Achituve, A

[Achituve et al., 2021] I. Achituve, A. Navon, Y . Yemini, G. Chechik, and E. Fetaya. Gp-tree: A gaussian process classifier for few-shot incremental learning. In ICML, pages 54–65. PMLR,

work page 2021
[5]

Boschini, L

[Boschini et al., 2022] M. Boschini, L. Bonicelli, P. Buzzega, A. Porrello, and S. Calderara. Class-incremental continual learn- ing into the extended der-verse. TPAMI, 45(5):5497–5512,

work page 2022
[6]

Chen and C.-G

[Chen and Lee, 2021] K. Chen and C.-G. Lee. Incremental few- shot learning via vector quantization in deep embedded space. In ICLR,

work page 2021
[7]

[Chi et al., 2022] Z. Chi, L. Gu, H. Liu, Y . Wang, Y . Yu, and J. Tang. Metafscil: A meta-learning approach for few-shot class incremental learning. In CVPR, pages 14166–14175,

work page 2022
[8]

[Cui et al., 2021] Y . Cui, W. Xiong, M. Tavakolian, and L. Liu. Semi-supervised few-shot class-incremental learning. In ICIP, pages 1239–1243. IEEE,

work page 2021
[9]

D ´ıaz-Rodr´ıguez, V

[D´ıaz-Rodr´ıguez et al., 2018] N. D ´ıaz-Rodr´ıguez, V . Lomonaco, D. Filliat, and D. Maltoni. Don’t forget, there is more than forgetting: new metrics for continual learning. arXiv preprint arXiv:1810.13166,

arXiv 2018
[12]

[Kang et al., 2022] H. Kang, J. Yoon, S. R. H. Madjid, S. J. Hwang, and C. D. Yoo. On the soft-subnetwork for few-shot class incre- mental learning. ICLR,

work page 2022
[13]

[Khan et al., 2023] M. G. Z. A. Khan, M. F. Naeem, L. Van Gool, D. Stricker, F. Tombari, and M. Z. Afzal. Introducing language guidance in prompt-based continual learning. In ICCV, pages 11463–11473,

work page 2023

Show all 54 references

[15]

[Kim et al., 2025] G. Kim, C. Xiao, T. Konishi, Z. Ke, and B. Liu. Open-world continual learning: Unifying novelty detection and continual learning. AI, 338:104237,

2025
[17]

[Li et al., 2023] Y . Li, W. Cao, W. Xie, J. Li, and E. Benetos. Few- shot class-incremental audio classification using dynamically ex- panded classifier with self-attention modified prototypes. TMM, 26:1346–1360,

2023
[18]

[Li et al., 2025] Y . Li, P. Moghadam, C. Peng, N. Ye, and P. Ko- niusz. Inductive graph few-shot class incremental learning. WSDM,

2025
[19]

Liang and W.-J

[Liang and Li, 2024] Y .-S. Liang and W.-J. Li. Inflora: Interference-free low-rank adaptation for continual learning. In CVPR, pages 23638–23647,

2024
[20]

[Lin et al., 2024] J. Lin, Z. Wu, W. Lin, J. Huang, and R. Luo. M2sd: Multiple mixing self-distillation for few-shot class- incremental learning. In AAAI, volume 38, pages 3422–3431,

2024
[21]

[Liu et al., 2020] J. Liu, L. Song, and Y . Qin. Prototype rectification for few-shot learning. In ECCV, pages 741–756. Springer,

2020
[23]

[Liu et al., 2023] B. Liu, B. Yang, L. Xie, R. Wang, Q. Tian, and Q. Ye. Learnable distribution calibration for few-shot class- incremental learning. TPAMI, 45(10):12699–12706,

2023
[24]

[Liu et al., 2024] C. Liu, Z. Wang, T. Xiong, R. Chen, Y . Wu, J. Guo, and H. Huang. Few-shot class incremental learning with attention-aware self-adaptive prompt. ECCV,

2024
[25]

[Lu et al., 2022] B. Lu, X. Gan, L. Yang, W. Zhang, L. Fu, and X. Wang. Geometer: Graph few-shot class-incremental learning via prototype representation. In KDD, pages 1152–1161,

2022
[26]

[Park et al., 2024] K.-H. Park, K. Song, and G.-M. Park. Pre- trained vision and language transformers are few-shot incremen- tal learners. In CVPR, pages 23881–23890,

2024
[27]

[Peng et al., 2022] C. Peng, K. Zhao, T. Wang, M. Li, and B. C. Lovell. Few-shot class-incremental learning from an open-set perspective. In ECCV, pages 382–397. Springer,

2022
[28]

Radford, J

[Radford et al., 2021] A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. In ICML, pages 8748–8763. PMLR,

2021
[29]

[Shi et al., 2021] G. Shi, J. Chen, W. Zhang, L.-M. Zhan, and X.-M. Wu. Overcoming catastrophic forgetting in incremental few-shot learning by finding flat minima. NIPS, 34:6747–6761,

2021
[30]

[Si et al., 2024] Y . Si, Y . Li, J. Li, J. Tan, and Q. He. Fully few-shot class-incremental audio classification using expandable dual-embedding extractor. Interspeech,

2024
[31]

[Smith et al., 2023] J. S. Smith, L. Karlinsky, V . Gutta, P. Cascante- Bonilla, D. Kim, A. Arbelle, R. Panda, R. Feris, and Z. Kira. Coda-prompt: Continual decomposed attention-based prompting for rehearsal-free continual learning. In CVPR, pages 11909– 11919,

2023
[32]

[Song et al., 2023] Z. Song, Y . Zhao, Y . Shi, P. Peng, L. Yuan, and Y . Tian. Learning with fantasy: Semantic-aware virtual con- trastive constraint for few-shot class-incremental learning. In CVPR, pages 24183–24192,

2023
[33]

[Tai et al., 2023] Y . Tai, Y . Tan, S. Xiong, and J. Tian. Mine-distill- prototypes for complete few-shot class-incremental learning in image classification. TGRS, 61:1–13,

2023
[34]

[Tan et al., 2022] Z. Tan, K. Ding, R. Guo, and H. Liu. Graph few- shot class-incremental learning. InWSDM, pages 987–996,

2022
[35]

Tang, Y .-X

[Tang et al., 2024] Y .-M. Tang, Y .-X. Peng, J. Meng, and W.-S. Zheng. Rethinking few-shot class-incremental learning: Learn- ing from yourself. ECCV,

2024
[36]

[Tao et al., 2020] X. Tao, X. Hong, X. Chang, S. Dong, X. Wei, and Y . Gong. Few-shot class-incremental learning. In CVPR, pages 12183–12192,

2020
[37]

[Tian et al., 2024] S. Tian, L. Li, W. Li, H. Ran, X. Ning, and P. Ti- wari. A survey on few-shot class-incremental learning. NN, 169:307–324,

2024
[38]

[Wang et al., 2021] Y . Wang, N. J. Bryan, M. Cartwright, J. P. Bello, and J. Salamon. Few-shot continual learning for audio classification. In ICASSP, pages 321–325. IEEE,

2021
[39]

[Wang et al., 2025] X. Wang, Z. Ji, X. Liu, Y . Pang, and J. Han. On the approximation risk of few-shot class-incremental learning. In ECCV, pages 162–178. Springer,

2025
[40]

[Xiao et al., 2023] J. Xiao, J. Li, and H. Gao. Fs3dciot: A few-shot incremental learning network for skin disease differential diag- nosis in the consumer iot. TCE, 69(4):786–797,

2023
[41]

Xinrui, C

[Xinrui et al., 2024] W. Xinrui, C. Geng, W. Wan, S.-Y . Li, and S. Chen. Forgetting, ignorance or myopia: Revisiting key chal- lenges in online continual learning. In NIPS,

2024
[42]

[Yang et al., 2021] B. Yang, M. Lin, B. Liu, M. Fu, C. Liu, R. Ji, and Q. Ye. Learnable expansion-and-compression net- work for few-shot class-incremental learning. arXiv preprint arXiv:2104.02281,

2021 arXiv
[43]

[Yang et al., 2022] B. Yang, M. Lin, Y . Zhang, B. Liu, X. Liang, R. Ji, and Q. Ye. Dynamic support network for few-shot class incremental learning. TPAMI, 45(3):2945–2951,

2022
[44]

[Yang et al., 2023a] H. Yang, W. Huang, J. Liu, C. Li, and S. Wang. Few-shot class-incremental learning for cross-domain disease classification. arXiv preprint arXiv:2304.05734,

arXiv
[45]

[Yang et al., 2023b] Y . Yang, H. Yuan, X. Li, Z. Lin, P. Torr, and D. Tao. Neural collapse inspired feature-classifier align- ment for few-shot class incremental learning. arXiv preprint arXiv:2302.03004,

arXiv
[46]

[Yoon et al., 2020] S. W. Yoon, D.-Y . Kim, J. Seo, and J. Moon. Xtarnet: Learning to extract task-adaptive representation for in- cremental few-shot learning. In ICML,

2020
[47]

Zhang, N

[Zhang et al., 2021] C. Zhang, N. Song, G. Lin, Y . Zheng, P. Pan, and Y . Xu. Few-shot incremental learning with continually evolved classifiers. In CVPR, pages 12455–12464,

2021
[48]

Zhang, P

[Zhang et al., 2024] J. Zhang, P. Zhao, Y . Zhao, C. Li, and D. Hu. Few-shot class-incremental learning for retinal disease recogni- tion. JBHI,

2024
[49]

[Zhao et al., 2021] H. Zhao, Y . Fu, M. Kang, Q. Tian, F. Wu, and X. Li. Mgsvf: Multi-grained slow versus fast framework for few- shot class-incremental learning. TPAMI, 46(3):1576–1588,

2021
[50]

[Zhao et al., 2022] Y . Zhao, L. Ha, H. Wang, and X. Ma. Few-shot class incremental learning for hyperspectral image classification based on constantly updated classifier. In IGARSS, pages 1376–

2022
[51]

[Zhu et al., 2021] K. Zhu, Y . Cao, W. Zhai, J. Cheng, and Z.-J. Zha. Self-promoted prototype refinement for few-shot class- incremental learning. In CVPR, pages 6801–6810,

2021
[52]

Zhuang, Z

[Zhuang et al., 2023] H. Zhuang, Z. Weng, R. He, Z. Lin, and Z. Zeng. Gkeal: Gaussian kernel embedded analytic learning for few-shot class incremental task. In CVPR,

2023
[53]

[Zou et al., 2022] Y . Zou, S. Zhang, Y . Li, and R. Li. Margin-based few-shot class-incremental learning with class-level overfitting mitigation. NIPS, 35:27267–27279,

2022
[54]

[Zou et al., 2024] Y . Zou, S. Zhang, H. Zhou, Y . Li, and R. Li. Compositional few-shot class-incremental learning. ICML, 2024

2024
[2017]

Hersche, G

[Hersche et al., 2022] M. Hersche, G. Karunaratne, G. Cherubini, L. Benini, A. Sebastian, and A. Rahimi. Constrained few-shot class-incremental learning. In CVPR, pages 9057–9067,

2022
[2018]

[Finn et al., 2017] C. Finn, P. Abbeel, and S. Levine. Model- agnostic meta-learning for fast adaptation of deep networks. In ICML, pages 1126–1135. PMLR,

2017
[2020]

[Liu et al., 2022] H. Liu, L. Gu, Z. Chi, Y . Wang, Y . Yu, J. Chen, and J. Tang. Few-shot class-incremental learning via entropy- regularized data-free replay. In ECCV, pages 146–162,

2022
[2021]

Ahmad, A

[Ahmad et al., 2022] T. Ahmad, A. R. Dhamija, S. Cruz, R. Rabi- nowitz, C. Li, M. Jafarzadeh, and T. E. Boult. Few-shot class in- cremental learning leveraging self-supervised features. In CVPR, pages 3900–3910,

2022
[2022]

Ahmed, A

[Ahmed et al., 2024] N. Ahmed, A. Kukleva, and B. Schiele. Orco: Towards better generalization via orthogonality and contrast for few-shot class-incremental learning. In CVPR, pages 28762– 28771,

2024
[2023]

Kim, D.-J

[Kim et al., 2023] D.-Y . Kim, D.-J. Han, J. Seo, and J. Moon. Warping the space: Weight space rotation for class-incremental few-shot learning. In ICLR,

2023
[2024]

[Aky¨urek et al., 2022] A. F. Aky¨urek, E. Aky ¨urek, D. Wijaya, and J. Andreas. Subspace regularizers for few-shot class incremental learning. In ICLR,

2022
[2025]

[Li and Liang, 2021] X. L. Li and P. Liang. Prefix-tuning: Op- timizing continuous prompts for generation. arXiv preprint arXiv:2101.00190,

2021 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [1]

Achituve, A

[Achituve et al., 2021] I. Achituve, A. Navon, Y . Yemini, G. Chechik, and E. Fetaya. Gp-tree: A gaussian process classifier for few-shot incremental learning. In ICML, pages 54–65. PMLR,

work page 2021

[2] [5]

Boschini, L

[Boschini et al., 2022] M. Boschini, L. Bonicelli, P. Buzzega, A. Porrello, and S. Calderara. Class-incremental continual learn- ing into the extended der-verse. TPAMI, 45(5):5497–5512,

work page 2022

[3] [6]

Chen and C.-G

[Chen and Lee, 2021] K. Chen and C.-G. Lee. Incremental few- shot learning via vector quantization in deep embedded space. In ICLR,

work page 2021

[4] [7]

[Chi et al., 2022] Z. Chi, L. Gu, H. Liu, Y . Wang, Y . Yu, and J. Tang. Metafscil: A meta-learning approach for few-shot class incremental learning. In CVPR, pages 14166–14175,

work page 2022

[5] [8]

[Cui et al., 2021] Y . Cui, W. Xiong, M. Tavakolian, and L. Liu. Semi-supervised few-shot class-incremental learning. In ICIP, pages 1239–1243. IEEE,

work page 2021

[6] [9]

D ´ıaz-Rodr´ıguez, V

[D´ıaz-Rodr´ıguez et al., 2018] N. D ´ıaz-Rodr´ıguez, V . Lomonaco, D. Filliat, and D. Maltoni. Don’t forget, there is more than forgetting: new metrics for continual learning. arXiv preprint arXiv:1810.13166,

arXiv 2018

[7] [12]

[Kang et al., 2022] H. Kang, J. Yoon, S. R. H. Madjid, S. J. Hwang, and C. D. Yoo. On the soft-subnetwork for few-shot class incre- mental learning. ICLR,

work page 2022

[8] [13]

[Khan et al., 2023] M. G. Z. A. Khan, M. F. Naeem, L. Van Gool, D. Stricker, F. Tombari, and M. Z. Afzal. Introducing language guidance in prompt-based continual learning. In ICCV, pages 11463–11473,

work page 2023