PEARL: Input-Agnostic Prompt Enhancement with Negative Feedback Regulation for Class-Incremental Learning

Hui Xue; Pengfei Fang; Yongchun Qin

REVIEW 4 major objections 5 minor 1 cited by

PEARL: Input-Agnostic Prompt Enhancement with Negative Feedback Regulation for Class-Incremental Learning

T0 review · 4 major / 5 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 PEARL claims input-agnostic prompts beat prior class-incremental methods by 2.24% average accuracy.

desk verdict A genuinely new prompt-based CIL method with strong numbers, but the negative-feedback story is under-validated and needs referee pressure. read the letter →

arxiv 2412.10900 v2 pith:3NQW5A3K submitted 2024-12-14 cs.LG cs.CV

Yongchun Qin , Pengfei Fang , Hui Xue This is my paper · ORCID

classification cs.LGcs.CV

keywords class-incrementallearningpre-trainedmodelspromptcatastrophicforgettinginput-agnosticnegativefeedbackregulationmomentumupdatecontinual

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 argues that prompt-based class-incremental learning has been held back by its reliance on input-dependent prompt selection, the "query-select" mechanism that mixes knowledge from different tasks in a disorganized way. It proposes PEARL, a method that replaces selection with an input-agnostic global prompt generated by a small prompt encoder, updated by a momentum rule whose weight is set by negative feedback from the model's own outputs. The claim is that this removes the model's dependence on data distribution, regulates forgetting without any distillation loss, and lets the prompt keep accumulating knowledge across sessions. If the claim holds, prompt-based methods move from lagging other pre-trained-model approaches to leading them, with state-of-the-art average accuracy on six benchmarks and a small memory footprint.

What carries the argument

Two coupled components carry the argument. Sequential Prompt Adaptation (SPA) treats the prompt pool as a sequence, assigns each task a frozen segment, and passes the concatenation through a session-sharing Transformer encoder to produce a global prompt token $[PT]^t_i$ with segmented positional encoding that marks which task each segment belongs to. Negative-feedback Knowledge Accumulation (NKA) then blends the previous session's prompt token with the current one via $[PT]^{mem}_i = \alpha_\tau [PT]^{t-1}_i + (1-\alpha_\tau)[PT]^t_i$, where $\alpha_\tau = \gamma \alpha_{\tau-1} + (1-\gamma)\sigma(\mathrm{MAE}(l_t[0:K(t-1)]\cdot\lambda, l_{t-1}\cdot\lambda))$. The MAE on the old-class slice is the feedback signal: it is the paper's proxy for how much old knowledge has been lost, and it makes the momentum weight adaptive rather than hand-tuned.

What would settle it

Take a stream whose new tasks come from a visually different domain and hold out a test set of previous-task classes. If PEARL's $\alpha_\tau$ is pushed up by the MAE on new-task inputs while true accuracy on the held-out old classes drops, or if PEARL with adaptive $\alpha$ does not beat a fixed $\alpha$ tuned on the same stream, the central claim is refuted. A direct version: measure the correlation between the NKA MAE and held-out old-class accuracy across sessions; a correlation near zero on any benchmark would show the feedback signal is not a forgetting proxy.

Watch

Extended reading notes

Core claim

PEARL's central claim is that catastrophic forgetting in pre-trained-model class-incremental learning can be controlled by a single global prompt that is the same for every input in a session, updated with an adaptive momentum weight. The global prompt is produced by a frozen-backbone prompt encoder over a task-partitioned prompt pool, so no per-input query or selection is needed. The momentum weight $\alpha_\tau$ is driven by the mean absolute error between current and previous logits restricted to old classes; high divergence raises $\alpha_\tau$ to preserve the old prompt, low divergence lowers it to admit new knowledge. The paper reports that this mechanism produces state-of-the-art results, beating the second-best method by 2.24% average accuracy and 1.65% final accuracy across six benchmarks, and that it converges to a task-dependent value independent of the initialization of $\alpha_0$.

Load-bearing premise

The whole adaptive schedule rests on the assumption that the mean absolute error between current and previous logits, measured only on old-class outputs for inputs of the current session, actually tracks how much old knowledge the model has forgotten; if new-task inputs are not informative about old classes, the feedback weight will not regulate forgetting correctly.

Editorial extensions

If this is right

Prompt-based class-incremental learning no longer needs input-dependent selection; inference on a new sample uses one global prompt token, cutting both compute and the risk of mismatched prompts.
The method is insensitive to the initial momentum weight: with fixed $\alpha$, accuracy collapses when $\alpha_0$ is low, while the NKA update maintains high accuracy across initial values from 0.6 to 0.999.
Knowledge retention is handled by the adaptive weight instead of a distillation loss, simplifying the training objective to plain cross-entropy.
The approach works across streams of different lengths and domains: six benchmarks from 5-task to 40-task streams all show average and final accuracy above the compared methods.

Reading between the lines

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

The NKA feedback signal is computed on current-task data, so if new-task inputs are far out-of-distribution for old classes, the MAE may stop tracking true forgetting; a testable extension is to replace the old-class slice with a small stored exemplar set and compare.
Because the paper shows $\alpha_\tau$ converges to a value that varies across tasks, the converged weight could be used as a measure of inter-task correlation to decide how many prompts to allocate to future tasks.
The same negative-feedback momentum recipe could transfer to domain-incremental or task-agnostic settings, where the old-class slice would be replaced by a domain-discrimination or novelty score.

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

Honest take: this is a real contribution. The combination of an input-agnostic global prompt—a small prompt encoder over a per-task partitioned pool with segmented positional encoding—and an adaptive momentum update is not something I've seen in L2P, DualPrompt, CODA-Prompt, OVOR, or HiDe-Prompt. The reported gains, about 2.2 average points over RanPAC across six benchmarks, are substantial if they hold. The ablations are mostly convincing: both the encoder and momentum are needed, and the adaptive momentum is robust to the initial α where fixed momentum collapses at low α. Code is available; the protocol is standard; the supplementary includes hyperparameter sweeps and forgetting rates.

The main soft spot is the NKA feedback signal. Eq. (10) computes MAE between current and previous logits on the current session's new-task images, restricted to old-class outputs. The paper calls that a measure of knowledge retention, but it never validates the proxy against actual old-task accuracy. On new-task images, old-class logits are out-of-distribution responses; their stability under a prompt update does not obviously track whether old-class accuracy has dropped. So the attractive story—negative feedback regulates forgetting—is not established. The ablation shows NKA beats fixed momentum, but that's consistent with the adaptive rule simply finding a good learning-rate schedule; Fig. 4 shows α converges to a constant for all initializations, which suggests the signal may be doing less diagnostic work than advertised.

Second, the SOTA numbers rest on hyperparameters (λ, γ, θ bounds, α0) selected from experiments on the same benchmarks used in the final comparison, with no proper validation split. That's common in this subfield, but it does weaken the significance claim. Also, Table 1 reports averages over three runs without error bars, so I can't tell how much of the 2.24% gap is noise.

Finally, the 'knowledge container theory' is a metaphor, not a theory, and 'first use of negative feedback in incremental learning' is the kind of claim that won't survive contact with the literature on control-theoretic continual learning. Those overclaims should be trimmed, but they don't undermine the method.

This deserves peer review. A referee should push for a direct test of the forgetting proxy: correlate mae with actual old-task accuracy drops, and add error bars. If the proxy validation fails, the paper still stands as an adaptive-momentum prompt method with strong numbers, just with a less exciting story.

Referee Report

4 major / 5 minor

Summary. The manuscript proposes PEARL, a rehearsal-free class-incremental learning method built on a frozen pretrained ViT. It removes the input-dependent "query-select" prompt choice and instead maintains an input-agnostic global prompt generated by a small prompt encoder over a prompt pool; a momentum update (Eq. 3) mixes the previous and current prompt tokens, and the momentum weight is adapted by the NKA mechanism (Eq. 11) using the MAE between current and previous logits on old-class outputs (Eq. 10). Experiments on six benchmarks and several ablations report average/final accuracy exceeding prior prompt-based and representation-based methods, including a 2.24% average-accuracy gain over RanPAC, the second-best method.

Significance. If the empirical claims hold, PEARL is a conceptually simple and potentially impactful result: it shows that an input-agnostic prompt with adaptive momentum can outperform prior prompt-based, representation-based, and model-mixture methods in PTM-based CIL, challenging the recent survey conclusion that prompt-based methods underperform. The paper ships code, uses standard benchmarks and backbones, reports multiple ablations including hyperparameter sensitivity, and includes an OVOR comparison and forgetting-rate tables; these are genuine strengths. The main weaknesses are that the forgetting-proxy signal in Eq. 10 is not validated against actual forgetting, hyperparameters are selected on the same benchmarks without a validation split, and Table 1 lacks variance estimates. These issues do not invalidate the method but weaken the causal-explanatory and state-of-the-art claims as currently stated.

major comments (4)

[Negative-feedback Knowledge Accumulation, Eq. (10); Table 4] The mae signal is computed on current-session images, using the old-class slice of the current logits and the previous logits for the same images. Since these images come from the new task, the old-class logits are responses to out-of-distribution inputs; stability of these responses under a prompt update does not by itself measure retention of previously learned classes. True forgetting is the drop in accuracy on old-task test data after learning a new task, and the paper never measures or correlates mae with such forgetting. The ablation in Table 4 and the convergence curves in Fig. 4 are consistent with NKA acting as a robust learning-rate schedule: alpha converges to a near-constant value regardless of initialization. Please add per-session measurements of actual forgetting on old-class test data, report a direct correlation between mae and forgetting, or include an ablation in which the adaptive signal is replaced by a fixed schedule to show that the adaptive component, not merely the converged schedule, is responsible for the reported gains.
[Sequential Prompt Adaptation, Eq. (7); Implementation Details] With the prompt-pool length M=100 and the ImageNet-R 40-task setting, M/N = 2.5, so the slice P[1 + M/N * (t-1) : M/N * t] is not an integer range. It is unclear whether the reported results use floor/ceil boundaries, an adjusted pool length, or some other allocation of prompts to tasks. Please specify the exact slicing used for each benchmark and ensure that the pseudo-code, Eq. (7), and the segmented positional encoding are consistent for all six benchmarks, since this directly affects reproducibility.
[Supplementary Section 5; Implementation Details] The values lambda=12500, gamma=0.99, theta_min/theta_max, and the prompt-pool length are selected experimentally on the same benchmarks used for the final comparison, with no held-out validation split or nested model-selection protocol. The supplementary text states that gamma and lambda are "determined experimentally" and reports tuning curves on VTAB and ImageNet-A. Tuning on the test benchmark can inflate the reported SOTA numbers. Please either use a validation split for hyperparameter selection or report the complete sensitivity grid and justify that the final choices do not materially change the comparison with RanPAC and other baselines.
[Table 1 and Comparison with State-of-the-art Methods] The text states that results are the average of three random runs, but Table 1 reports no standard deviations or confidence intervals. Given that the gain over RanPAC is only 0.13% in average accuracy on CIFAR-100 and PEARL's final accuracy is lower than RanPAC there (89.02 vs. 89.30), the "state-of-the-art" claim needs error bars and, ideally, a paired comparison across the three runs to establish that the reported improvements are not within run-to-run variation.

minor comments (5)

[Eq. (4)] The objective is written as arg max E_{(x,y)~D_t} I(y != l_t), which maximizes the error; it should be I(y = l_t) or an equivalent loss-minimization form.
[Datasets, Experiments] The text refers to "CIFAR00" in the datasets paragraph; this should be CIFAR-100.
[Eq. (10)] Please specify the shapes of l_t and l_{t-1} and clarify where lambda is applied; as written it is ambiguous whether lambda multiplies both terms and how the different lengths of the two logit vectors are handled.
[Supplementary Table 4] The method name "RacPAC" is a typo; it should be RanPAC.
[Algorithm 1] Line 10 contains the typo "Initilize"; also, the pseudo-code resets alpha_tau to 0.99 at the start of every session, which should be stated explicitly in the main text alongside Eq. (11).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PEARL's SPA and NKA are empirical designs with operational equations; the unvalidated mae proxy and same-benchmark hyperparameter tuning are correctness risks, not circular reductions.

full rationale

The paper does not claim a formal derivation from first principles; it presents an empirical method and reports measured accuracies. Eq. (3) defines the momentum prompt as a convex combination of previous and current prompt tokens, Eq. (10) defines the mae feedback signal as a divergence between current and previous logits, and Eq. (11) updates the momentum weight from that signal. These are operational definitions of a feedback control loop, not a derivation in which the claimed performance is presupposed. The statement that low mae indicates good knowledge retention is an interpretation of the proxy; if the proxy fails to track true forgetting, that is a validity concern, not circularity, because no equation reduces old-task test accuracy to mae. Similarly, the supplementary selection of hyperparameters such as lambda, gamma, and the sigmoid bounds on the same six benchmarks is an experimental overfitting risk, not a constructional circularity: the final numbers are measured accuracies rather than predictions forced by the fitted constants. There are no load-bearing self-citations: the authors do not invoke a uniqueness theorem or prior result of their own to justify the method, and citations to CODA-Prompt, RanPAC, and other works are standard baselines or design choices. No circular step can be exhibited with a specific quote and reduction, so the appropriate finding is no significant circularity.

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

The central SOTA claim depends on several hand-tuned hyperparameters and on unproven modeling assumptions: that a single global prompt can encode cross-task knowledge, that the MAE-based feedback measures retention, and that the converged alpha reflects an 'inherent task correlation'. No new physical entities are introduced.

free parameters (5)

lambda = 12500
Scale factor in MAE (Eq. 10); tuned experimentally, no validation split described.
gamma = 0.99
Momentum rate for alpha update (Eq. 11); chosen because performance improves as gamma increases, per Supplementary Fig. 4.
theta_min and theta_max = 0.7 and 0.999
Bounds of the sigmoid-like activation; chosen to keep alpha stable, per Supplementary Fig. 5.
alpha_0 = 0.99
Initial momentum weight in NKA; chosen with a bias toward old knowledge.
Prompt pool length M, encoder depth L, prompt token length H = M=100, L=2, H=4
Selected via ablations (Tables 5-6, Fig. 3) to balance performance and cost.

assumptions (4)

domain assumption Frozen ViT-B/16 provides a sufficiently strong feature extractor for all downstream CIL tasks.
The entire method keeps V(.) frozen and only learns prompts; if the pretrained features are poor for a domain, the method inherits that weakness.
ad hoc to paper The prompt encoder, a 2-layer transformer, can capture cross-task correlations by attending to a sequence of prompt slices.
This is the core design of SPA, stated in Methodology, but no independent evidence is given that such correlations are real or that this architecture captures them.
ad hoc to paper MAE between current and previous logits on the old-class slice of current-session data measures knowledge retention.
This is the load-bearing assumption of NKA; the paper provides no validation that this proxy tracks forgetting on old-class data.
domain assumption The RanPAC classifier, which uses random projections and a closed-form head, is an appropriate classifier under this prompt scheme.
Borrowed from McDonnell et al.; PEARL contributes no analysis of head behavior under its prompt updates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of PEARL: Input-Agnostic Prompt Enhancement with Negative Feedback Regulation for Class-Incremental Learning." pith.science (2026). https://pith.science/paper/3NQW5A3K

@misc{pith2026241210900,
  author       = {Pith},
  title        = {Pith review of: PEARL: Input-Agnostic Prompt Enhancement with Negative Feedback Regulation for Class-Incremental Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3NQW5A3K}},
  note         = {Machine review of arXiv:2412.10900}
}

read the original abstract

Class-incremental learning (CIL) aims to continuously introduce novel categories into a classification system without forgetting previously learned ones, thus adapting to evolving data distributions. Researchers are currently focusing on leveraging the rich semantic information of pre-trained models (PTMs) in CIL tasks. Prompt learning has been adopted in CIL for its ability to adjust data distribution to better align with pre-trained knowledge. This paper critically examines the limitations of existing methods from the perspective of prompt learning, which heavily rely on input information. To address this issue, we propose a novel PTM-based CIL method called Input-Agnostic Prompt Enhancement with Negative Feedback Regulation (PEARL). In PEARL, we implement an input-agnostic global prompt coupled with an adaptive momentum update strategy to reduce the model's dependency on data distribution, thereby effectively mitigating catastrophic forgetting. Guided by negative feedback regulation, this adaptive momentum update addresses the parameter sensitivity inherent in fixed-weight momentum updates. Furthermore, it fosters the continuous enhancement of the prompt for new tasks by harnessing correlations between different tasks in CIL. Experiments on six benchmarks demonstrate that our method achieves state-of-the-art performance. The code is available at: https://github.com/qinyongchun/PEARL.

Figures

Figures reproduced from arXiv: 2412.10900 by the authors.

**Figure 2.** The illustration of the proposed PEARL. The ViT consists of 12 blocks, with a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** The illustration of the proposed NKA mechanism. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: The curves of mae and α, across different initial value of α 0 . Results are derived from the second session of CUB. α 0 Fixed α NKA α A A ¯ N A A ¯ N 0.60 60.04 44.83 81.56 73.88 0.70 61.41 47.96 84.47 76.89 0.80 68.79 63.74 87.28 79.90 0.90 86.32 82.15 89.73 83.40 0.…

**Figure 5.** Figure 5: The value of α final across different settings. Conclusion This paper presents PEARL, an input-agnostic prompt method designed to address the issue of knowledge interference caused by the “query-select” mechanism of existing input-dependent prompt methods. Our method…

**Figure 1.** Figure 1: The comparison between CPE and SPE, where [PITH_FULL_IMAGE:figures/full_fig_p011_1.png]

**Figure 3.** Figure 3: Ablation study on prompt encoder depth. 2 4 6 8 10 Value of 85 90 95 Averave Accuracy 1 10 100 1000 1e4 1e5 5·1e5 1e6 1.5·1e6 2·1e6 0.90 0.92 0.94 0.96 0.98 Value of 85 90 95 Averave Accuracy [PITH_FULL_IMAGE:figures/full_fig_p013_3.png]

**Figure 4.** Figure 4: Ablation study on λ and γ. 5 Further Ablation Study In this section, we present further ablation study, focusing on hyperparameter selection. For the proposed SPA module, H, the length of prompt token and L, the depth of prompt encoder are undetermined hyperparameters …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Forward-Only Continual Learning
cs.LG 2025-09 conditional novelty 6.0 of 10

FoRo achieves strong continual learning accuracy and low forgetting on CIFAR-100, ImageNet-R, and CUB-200 using only forward updates, via CMA-ES prompt tuning and a recursive knowledge encoding matrix.

Reference graph

Works this paper leans on

50 extracted references · 16 canonical work pages · cited by 1 Pith paper

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Aljundi, R.; Babiloni, F.; Elhoseiny, M.; Rohrbach, M.; and Tuytelaars, T. 2018. Memory aware synapses: Learning what (not) to forget. In Proceedings of the European conference on computer vision (ECCV), 139--154

work page 2018
[4]

M.; Mar \' n-Jim \'e nez, M

Castro, F. M.; Mar \' n-Jim \'e nez, M. J.; Guil, N.; Schmid, C.; and Alahari, K. 2018 a . End-to-end incremental learning. In Proceedings of the European Conference on Computer Vision, 233--248

work page 2018
[5]

M.; Mar \' n-Jim \'e nez, M

Castro, F. M.; Mar \' n-Jim \'e nez, M. J.; Guil, N.; Schmid, C.; and Alahari, K. 2018 b . End-to-end incremental learning. In Proceedings of the European conference on computer vision (ECCV), 233--248

work page 2018
[6]

K.; Ajanthan, T.; and Torr, P

Chaudhry, A.; Dokania, P. K.; Ajanthan, T.; and Torr, P. H. 2018. Riemannian walk for incremental learning: Understanding forgetting and intransigence. In Proceedings of the European conference on computer vision (ECCV), 532--547

2018
[7]

Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

arXiv 2020
[8]

Gao, Z.; Cen, J.; and Chang, X. 2024. Consistent Prompting for Rehearsal-Free Continual Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 28463--28473

work page 2024

Show all 50 references

[9]

Hendrycks, D.; Basart, S.; Mu, N.; Kadavath, S.; Wang, F.; Dorundo, E.; Desai, R.; Zhu, T.; Parajuli, S.; Guo, M.; et al. 2021 a . The many faces of robustness: A critical analysis of out-of-distribution generalization. In Proceedings of the IEEE/CVF international conference o...

2021
[10]

Hendrycks, D.; Zhao, K.; Basart, S.; Steinhardt, J.; and Song, D. 2021 b . Natural adversarial examples. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 15262--15271

2021
[11]

Hinton, G.; Vinyals, O.; and Dean, J. 2015. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531

2015 arXiv
[12]

C.; Wang, Z.; and Lin, D

Hou, S.; Pan, X.; Loy, C. C.; Wang, Z.; and Lin, D. 2019 a . Learning a unified classifier incrementally via rebalancing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 831--839

2019
[13]

C.; Wang, Z.; and Lin, D

Hou, S.; Pan, X.; Loy, C. C.; Wang, Z.; and Lin, D. 2019 b . Learning a unified classifier incrementally via rebalancing. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 831--839

2019
[14]

J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685

2021 arXiv
[15]

Huang, W.-C.; Chen, C.-F.; and Hsu, H. 2024. OVOR: OnePrompt with Virtual Outlier Regularization for Rehearsal-Free Class-Incremental Learning. arXiv preprint arXiv:2402.04129

2024 arXiv
[16]

Jia, M.; Tang, L.; Chen, B.-C.; Cardie, C.; Belongie, S.; Hariharan, B.; and Lim, S.-N. 2022. Visual prompt tuning. In European Conference on Computer Vision, 709--727. Springer

2022
[17]

Jung, H.; Ju, J.; Jung, M.; and Kim, J. 2016. Less-forgetting learning in deep neural networks. arXiv preprint arXiv:1607.00122

2016 arXiv
[18]

Kim, Y.; Fang, J.; Zhang, Q.; Cai, Z.; Shen, Y.; Duggal, R.; S Raychaudhuri, D.; Tu, Z.; Xing, Y.; and Dabeer, O. 2025. Open-World Dynamic Prompt and Continual Visual Representation Learning. In European Conference on Computer Vision, 357--374. Springer

2025
[19]

A.; Milan, K.; Quan, J.; Ramalho, T.; Grabska-Barwinska, A.; et al

Kirkpatrick, J.; Pascanu, R.; Rabinowitz, N.; Veness, J.; Desjardins, G.; Rusu, A. A.; Milan, K.; Quan, J.; Ramalho, T.; Grabska-Barwinska, A.; et al. 2017. Overcoming catastrophic forgetting in neural networks. Proceedings of the national academy of sciences, 114(13): 3521--3526

2017
[20]

Krizhevsky, A.; Hinton, G.; et al. 2009. Learning multiple layers of features from tiny images

2009
[21]

Lester, B.; Al-Rfou, R.; and Constant, N. 2021. The power of scale for parameter-efficient prompt tuning. arXiv preprint arXiv:2104.08691

2021 arXiv
[22]

L.; and Liang, P

Li, X. L.; and Liang, P. 2021. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190

2021 arXiv
[23]

Li, Z.; and Hoiem, D. 2017. Learning without forgetting. IEEE transactions on pattern analysis and machine intelligence, 40(12): 2935--2947

2017
[24]

D.; and Van De Weijer, J

Masana, M.; Liu, X.; Twardowski, B.; Menta, M.; Bagdanov, A. D.; and Van De Weijer, J. 2022. Class-incremental learning: survey and performance evaluation on image classification. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(5): 5513--5533

2022
[25]

McCloskey, M.; and Cohen, N. J. 1989. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation, volume 24, 109--165. Elsevier

1989
[26]

D.; Gong, D.; Parvaneh, A.; Abbasnejad, E.; and van den Hengel, A

McDonnell, M. D.; Gong, D.; Parvaneh, A.; Abbasnejad, E.; and van den Hengel, A. 2024. Ranpac: Random projections and pre-trained models for continual learning. Advances in Neural Information Processing Systems, 36

2024
[27]

Ostapenko, O.; Puscas, M.; Klein, T.; Jahnichen, P.; and Nabi, M. 2019. Learning to remember: A synaptic plasticity driven framework for continual learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11321--11329

2019
[28]

Paszke, A.; Gross, S.; Massa, F.; Lerer, A.; Bradbury, J.; Chanan, G.; Killeen, T.; Lin, Z.; Gimelshein, N.; Antiga, L.; et al. 2019. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32

2019
[29]

Rebuffi, S.-A.; Kolesnikov, A.; Sperl, G.; and Lampert, C. H. 2017 a . i C a RL : Incremental classifier and representation learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2001--2010

2017
[30]

Rebuffi, S.-A.; Kolesnikov, A.; Sperl, G.; and Lampert, C. H. 2017 b . icarl: Incremental classifier and representation learning. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, 2001--2010

2017
[31]

K.; Kim, J.; and Kim, J

Shin, H.; Lee, J. K.; Kim, J.; and Kim, J. 2017. Continual learning with deep generative replay. Advances in neural information processing systems, 30

2017
[32]

S.; Karlinsky, L.; Gutta, V.; Cascante-Bonilla, P.; Kim, D.; Arbelle, A.; Panda, R.; Feris, R.; and Kira, Z

Smith, J. S.; Karlinsky, L.; Gutta, V.; Cascante-Bonilla, P.; Kim, D.; Arbelle, A.; Panda, R.; Feris, R.; and Kira, Z. 2023. Coda-prompt: Continual decomposed attention-based prompting for rehearsal-free continual learning. In Proceedings of the IEEE/CVF Conference on Computer...

2023
[33]

Sun, H.-L.; Zhou, D.-W.; Ye, H.-J.; and Zhan, D.-C. 2023. PILOT: A Pre-Trained Model-Based Continual Learning Toolbox. arXiv preprint arXiv:2309.07117

2023 arXiv
[34]

N.; Kaiser, .; and Polosukhin, I

Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is all you need. Advances in neural information processing systems, 30

2017
[35]

Wah, C.; Branson, S.; Welinder, P.; Perona, P.; and Belongie, S. 2011. The caltech-ucsd birds-200-2011 dataset

2011
[36]

Wang, L.; Xie, J.; Zhang, X.; Huang, M.; Su, H.; and Zhu, J. 2024. Hierarchical decomposition of prompt-based continual learning: Rethinking obscured sub-optimality. Advances in Neural Information Processing Systems, 36

2024
[37]

Wang, Y.; Huang, Z.; and Hong, X. 2022. S-prompts learning with pre-trained transformers: An occam’s razor for domain incremental learning. Advances in Neural Information Processing Systems, 35: 5682--5695

2022
[38]

Wang, Y.; Ma, Z.; Huang, Z.; Wang, Y.; Su, Z.; and Hong, X. 2023. Isolation and impartial aggregation: A paradigm of incremental learning without interference. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 10209--10217

2023
[39]

Wang, Z.; Zhang, Z.; Ebrahimi, S.; Sun, R.; Zhang, H.; Lee, C.-Y.; Ren, X.; Su, G.; Perot, V.; Dy, J.; et al. 2022 a . Dualprompt: Complementary prompting for rehearsal-free continual learning. In European Conference on Computer Vision, 631--648. Springer

2022
[40]

Wang, Z.; Zhang, Z.; Lee, C.-Y.; Zhang, H.; Sun, R.; Ren, X.; Su, G.; Perot, V.; Dy, J.; and Pfister, T. 2022 b . Learning to prompt for continual learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 139--149

2022
[41]

Wu, Y.; Chen, Y.; Wang, L.; Ye, Y.; Liu, Z.; Guo, Y.; and Fu, Y. 2019. Large scale incremental learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 374--382

2019
[42]

Xiang, Y.; Fu, Y.; Ji, P.; and Huang, H. 2019. Incremental learning using conditional adversarial networks. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 6619--6628

2019
[43]

B.; Ravfogel, S.; and Goldberg, Y

Zaken, E. B.; Ravfogel, S.; and Goldberg, Y. 2021. Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models. arXiv preprint arXiv:2106.10199

2021
[44]

S.; Neumann, M.; Dosovitskiy, A.; et al

Zhai, X.; Puigcerver, J.; Kolesnikov, A.; Ruyssen, P.; Riquelme, C.; Lucic, M.; Djolonga, J.; Pinto, A. S.; Neumann, M.; Dosovitskiy, A.; et al. 2019. A large-scale study of representation learning with the visual task adaptation benchmark. arXiv preprint arXiv:1910.04867

2019 arXiv
[45]

Zhang, C.; Zhang, L.; Wu, J.; Zhou, D.; and He, Y. 2024. Causal prompting: Debiasing large language model prompting based on front-door adjustment. arXiv preprint arXiv:2403.02738

2024 arXiv
[46]

Zhang, Y.; Yin, Z.; Shao, J.; and Liu, Z. 2022. Benchmarking omni-vision representation through the lens of visual realms. In European Conference on Computer Vision, 594--611. Springer

2022
[47]

Zhou, D.-W.; Sun, H.-L.; Ning, J.; Ye, H.-J.; and Zhan, D.-C. 2024 a . Continual learning with pre-trained models: A survey. arXiv preprint arXiv:2401.16386

2024 arXiv
[48]

Zhou, D.-W.; Sun, H.-L.; Ye, H.-J.; and Zhan, D.-C. 2024 b . Expandable subspace ensemble for pre-trained model-based class-incremental learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 23554--23564

2024
[49]

Zhou, D.-W.; Wang, Q.-W.; Qi, Z.-H.; Ye, H.-J.; Zhan, D.-C.; and Liu, Z. 2024 c . Class-Incremental Learning: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence

2024
[50]

Zhou, D.-W.; Ye, H.-J.; Zhan, D.-C.; and Liu, Z. 2023. Revisiting class-incremental learning with pre-trained models: Generalizability and adaptivity are all you need. arXiv preprint arXiv:2303.07338

2023 arXiv

Pith tools

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

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Aljundi, R.; Babiloni, F.; Elhoseiny, M.; Rohrbach, M.; and Tuytelaars, T. 2018. Memory aware synapses: Learning what (not) to forget. In Proceedings of the European conference on computer vision (ECCV), 139--154

work page 2018

[4] [4]

M.; Mar \' n-Jim \'e nez, M

Castro, F. M.; Mar \' n-Jim \'e nez, M. J.; Guil, N.; Schmid, C.; and Alahari, K. 2018 a . End-to-end incremental learning. In Proceedings of the European Conference on Computer Vision, 233--248

work page 2018

[5] [5]

M.; Mar \' n-Jim \'e nez, M

Castro, F. M.; Mar \' n-Jim \'e nez, M. J.; Guil, N.; Schmid, C.; and Alahari, K. 2018 b . End-to-end incremental learning. In Proceedings of the European conference on computer vision (ECCV), 233--248

work page 2018

[6] [6]

K.; Ajanthan, T.; and Torr, P

Chaudhry, A.; Dokania, P. K.; Ajanthan, T.; and Torr, P. H. 2018. Riemannian walk for incremental learning: Understanding forgetting and intransigence. In Proceedings of the European conference on computer vision (ECCV), 532--547

2018

[7] [7]

Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

arXiv 2020

[8] [8]

Gao, Z.; Cen, J.; and Chang, X. 2024. Consistent Prompting for Rehearsal-Free Continual Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 28463--28473

work page 2024