Pith. sign in

REVIEW 3 major objections 6 minor 78 references

Continual Multiple Instance Learning with Enhanced Localization for Histopathological Whole Slide Image Analysis

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Continual multiple instance learning on whole-slide images can keep both bag-level accuracy and pixel-level tumor localization while a model adapts to a sequence of new organs and subtypes, and CoMEL claims to achieve this rehearsal-free.

desk verdict Solid, thoroughly benchmarked continual MIL paper where the localization headline rests on self-training pseudo-labels that deserve independent verification before you trust the IoU gains. read the letter →

arxiv 2507.02395 v2 pith:PKI33Q3F submitted 2025-07-03 cs.CV

Byung Hyun Lee , Wongi Jeong , Woojae Han , Kyoungbun Lee , Se Young Chun This is my paper · ORCID
classification cs.CV
keywords continuallearningmultipleinstancewholeslideimageanalysishistopathologylocalizationpseudo-labelinglow-rankadaptationcatastrophicforgetting
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 goes after an under-explored problem: multiple instance learning (MIL) on histopathological whole slide images is cheap in annotation (only slide-level labels) but is almost always trained once, whereas real hospitals accumulate slides of new organs and tumor subtypes continuously. The authors propose CoMEL, a framework for continual MIL that must keep predicting the right slide class and the right tumor location after every new task. It does so with a grouped double-attention transformer for scaling to roughly $10^{5}$ patches per slide, bag-prototype pseudo-labeling that turns attention into instance-level training signal, and an orthogonal low-rank adaptation that adds per-task subspaces without overwriting earlier ones. On a five-organ sequence (CAMELYON-16 plus PAIP), CoMEL reports the best instance accuracy, IoU, Dice, and bag accuracy among all baselines, and on four TCGA tumor-subtype tasks it reports the lowest bag-level forgetting; if these results hold, continual re-training on hospital data becomes a realistic way to keep pathology models current without storing patient patches.

What carries the argument

The argument is carried by two mechanisms working together. Bag Prototypes-based Pseudo-Labeling (BPPL) converts the MIL attention vector $A$ into a class-probability vector by min-max normalization, builds a positive prototype $x_{\mathrm{bag}} = \sum_m a_m x_m$ and a negative prototype from the reversed attention $\tilde{a}_m = (1-\alpha_m)/\sum_m (1-\alpha_m)$, and pseudo-labels instances by their residual similarity to the two prototypes, discarding labels whenever the bag prediction is wrong or confidence is below the adaptive threshold $\tau_1$. Orthogonal Weighted Low-Rank Adaptation (OWLoRA) represents each task's weights as a sum of low-rank bases with orthonormal columns, imposing intra-orthogonality through a penalty $L_{\mathrm{in}} = \|U_n^T U_n - I_d\|_F^2$ and inter-orthogonality through gradient projection $\hat{G}(U_n) = (I_D - \sum_{k<n} U_k U_k^T)\, G(U_n)$; because the first task's principal components are frozen, later tasks literally cannot move the subspace the first task occupies. Grouped Double Attention Transformer (GDAT) makes these trainable and scalable: it rewrites the $O(M^2)$ attention as two $O(mM)$ attentions over average-pooled region tokens, then adds the raw values back with weight $\eta$ so grouped attention does not flatten instance diversity.

What would settle it

Two checks would settle the central claim. First, run CoMEL on a ten-task organ sequence and plot task-1 localization IoU after each new task: OWLoRA's orthogonal basis budget is finite, so if the mechanism is the reason forgetting stays low, forgetting should creep up as the subspace fills. Second, train the first task with a randomly frozen attention head and measure whether BPPL's pseudo-labels and bag-prototype losses still raise IoU over that poor-attention baseline; if IoU stays flat, the method's localization gains depend on initially good attention, exactly the assumption stated in Sec. 3.3.

Watch

Extended reading notes

Core claim

The central claim is that under a continual MIL setup, a model can keep learning new whole slide image tasks (new organs, or new tumor subtypes within an organ) while preserving both the slide-level classification and the instance-level localization that existing continual MIL methods mostly neglect. CoMEL's first task trains full weights, extracts the top principal components by singular value decomposition, and then freezes them; every subsequent task adds low-rank bases $U_k S_k (V_k)^T$ that are forced to be orthonormal and whose gradients are projected onto the orthogonal complement of all earlier bases, so new knowledge cannot displace old. Localization is driven by turning attention scores into per-instance class probabilities, forming positive and negative bag prototypes, and pseudo-labeling instances whose confidence passes an adaptive threshold gated on the bag prediction. Against the prior LoRA-based state of the art (InfLoRA) on the combined organ sequence, CoMEL reports ACCinst 72.64 versus 70.17, IoU 41.87 versus 31.87, Dice 51.44 versus 40.84, and ACCbag 62.96 versus 56.93; on the TCGA subtype sequence it reports ACCbag 87.94, Forgetbag 10.12, and M.ACCbag 92.28, the best of all compared methods.

Load-bearing premise

The load-bearing assumption is that the MIL model's attention already points at positive instances reasonably well: BPPL builds its prototypes and pseudo-labels from that attention, so if attention is poor at the start, the model trains on its own noisy labels and the claimed localization gains would not appear.

Editorial extensions

If this is right

  • Rehearsal-free continual MIL can beat rehearsal-based methods on bag accuracy while storing no patient patches, addressing the privacy concern that limits memory-based approaches like ER and ConSlide in hospitals.
  • Instance-level forgetting under continual MIL is substantially reducible: CoMEL reports Forgetinst 14.00 on the five-organ sequence versus 17.89 for InfLoRA and roughly 19-27 for the other baselines, with the pattern reproduced under reversed task order in Table S13.
  • The localization machinery also helps outside the continual setup: on the merged five-organ data, GDAT+BPPL raises instance accuracy to 79.27, IoU to 49.32, and Dice to 60.68, versus 68.24, 36.45, and 49.33 for the strongest localization baseline smAP.
  • CoMEL is backbone-agnostic: with ResNet50, PLIP, and CONCH as frozen feature extractors it still beats the continual baselines on Dice, showing the gains do not depend on the UNI foundation model.
  • On TCGA tumor subtyping, continual bag classification reaches ACCbag 87.94 and Forgetbag 10.12, leaving a small gap to the joint-training upper bound of ACCbag 91.49, which suggests the continual setting can approach the non-continual ceiling.

Reading between the lines

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

  • Making BPPL's pseudo-labels conditional on the model's own attention (the paper explicitly assumes the MIL model can successfully localize positive instances in Sec. 3.3) implies the method's gains should shrink for backbones whose first-task attention is poor; measuring per-task pseudo-label accuracy against attention quality would locate where the real dependency on pre-trained features sits.
  • OWLoRA's orthogonal subspace has a finite rank budget (the paper concedes this in Sec. S5), so on open-ended sequences of ten or more tasks the forgetting curve should eventually rise; a long-sequence experiment would settle whether the mechanism scales beyond the five-task setups tested here.
  • The positive-versus-reversed-attention prototype contrast is a general mechanism: the same idea could be dropped into other weakly supervised continual settings, such as incremental semantic segmentation from image-level labels, where pseudo-label noise similarly accumulates across tasks.
  • The paper leaves blurry tasks (mixed organs within a single task, which is what hospital data streams actually look like) as future work; whether BPPL's adaptive threshold and OWLoRA's task-specific bases survive ambiguous task boundaries is an open empirical question.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes CoMEL, a continual multiple instance learning (MIL) framework for whole slide image analysis that jointly addresses bag-level classification and instance-level localization. It introduces three components: Grouped Double Attention Transformer (GDAT) for scalable instance feature refinement, Bag Prototypes-based Pseudo-Labeling (BPPL) for generating instance pseudo-labels from attention and bag prototypes, and Orthogonal Weighted Low-Rank Adaptation (OWLoRA) for rehearsal-free continual learning with orthogonal low-rank bases. The method is evaluated on sequential organ datasets from CAMELYON-16 and PAIP for tumor detection and on TCGA for tumor subtyping. The main empirical claims are that CoMEL outperforms prior continual learning baselines in both bag-level accuracy and localization metrics (e.g., ACCinst 72.64, IoU 41.87, Dice 51.44, ACCbag 62.96 in Table 1) while incurring the least forgetting. The supplementary material includes ablations, efficiency analyses, reversed task order experiments, and an explicit limitation statement about BPPL's reliance on attention quality.

Significance. If the reported results hold, CoMEL fills a genuine gap: continual learning for MIL has mostly focused on bag classification, whereas localization under non-stationary tasks is clinically important and relatively unexplored. The evaluation is extensive, with 10-run comparisons in two task orders, ablations of all major components, and robustness checks across feature extractors. The pipeline is rehearsal-free, which is attractive for privacy-sensitive medical deployments, and OWLoRA is described clearly. The main uncertainty is whether BPPL's self-training loop is reliable beyond the tuned setting; this is a load-bearing issue for the paper's central claim of "enhanced localization" and is acknowledged in the paper's own Sec. S5.

major comments (3)
  1. [Sec. 3.3, Eqs. (5)-(9)] The localization gain attributed to BPPL rests on a self-training loop: the pseudo-labels in Eq. (7) are derived from the attention A and the prototypes x_bag and x_tilde_bag, and Eq. (8) then trains the same attention mechanism on those pseudo-labels, filtered only by bag correctness and confidence. The paper's own Sec. S5 states that BPPL "heavily relies on the accuracy of attention scores as pseudo-labels," and Tables S7-S8 show high sensitivity to T, tau, lambda1, and lambda2. The current evidence in Fig. S1(a) reports pseudo-label accuracy during training but gives no initial value and no comparison against a no-BPPL control, so it cannot rule out that the gains in Table 3 (ACCinst 79.27 vs. 68.36, IoU 49.32 vs. 36.31) are driven by favorable initialization plus hyperparameter selection. Please add: (i) pseudo-label accuracy and its agreement with ground truth at task start and at each task boundary; (ii) an ablation that applies the same filtering and loss weighting to attention scores without the prototype-based re-labeling; and (iii) an analysis of how performance changes when initial attention is deliberately degraded, to test whether BPPL corrects or amplifies errors. These analyses are necessary to support the claim of "reliable instance pseudo-labeling."
  2. [Sec. 4.1 and Table S1] The negative bags for the PAIP organs are constructed by randomly splitting tumor slides into two halves and deleting all tumor regions from the "normal" half. Thus, the normal bags are not independent normal WSIs but normal tissue extracted from cancer patients, and the binary tumor/normal split within each organ is artificial. The reported IoU and ACCinst for the PAIP tasks in Table 1 may therefore reflect this synthetic distribution rather than a realistic deployment distribution. Please clarify whether the official PAIP challenge provides truly normal slides, report results on them if available, and otherwise discuss this as an explicit limitation in the main text. This is load-bearing because the combined CM-16+PAIP sequence is the primary evidence for continual instance localization.
  3. [Sec. 4.1 and Tables S7-S8] The implementation section states that the same hyperparameters are used for all tasks, but no validation protocol is described. The ablations show that ACCinst ranges from 60.32 to 74.15 as T varies, from 61.23 to 74.15 as tau varies, and from 61.23 to 74.15 as lambda1 varies (Tables S7-S8), yet the chosen values are reported only as final settings. Because several baselines in Table 1 use the same GDAT+BPPL backbone, the margin of CoMEL over InfLoRA (IoU 41.87 vs. 31.87) might partly reflect hyperparameter tuning rather than the continual-learning mechanism. Please specify how (T, tau, lambda1, lambda2, eta, epsilon, d, lambda3) were selected, for example using a held-out validation set disjoint from the test slides, and provide a robustness table over a range of reasonable values for the main continual sequence.
minor comments (6)
  1. [Sec. S1.2.2, Eq. (S5)] The equation labeled Forgetinst is defined using Rbag_{n,l} and is used for bag-level forgetting; it should be labeled Forgetbag. The typo is confusing because the same symbol is also used for instance-level forgetting in Eq. (S3).
  2. [Sec. 3.2, Eq. (3)] It would clarify what "the overall complexity of Eq. (3) is also O(mM)" means in terms of the group size F; the current sentence is easy to misread as saying the complexity is independent of M. Please state explicitly that the complexity is linear in M for fixed m, and define m in terms of the grouping window F.
  3. [Table 5 and Sec. 3.4] The ablation column header G_hat(V_n; E) is not defined in the main text; Eq. (11) describes gradient projection for U_n only. Please specify how projection is applied to V_n and to the aggregator E, or remove the undefined notation.
  4. [Sec. 4.6] The sentence "ablating GDAT, leading to the results of ABMIL" is imprecise: ABMIL is a separate architecture, not GDAT with the grouped attention removed. The supplementary comparison with "Single Attn" in Table S11 is the right kind of ablation; the main text should refer to it directly.
  5. [Fig. S1(a)] The pseudo-label accuracy curve lacks axis labels and error bars and does not indicate which task it is averaged over. Please add these details and report the first-epoch value, which is needed to assess the starting point of the self-training loop.
  6. [Abstract] The claimed "up to 23.4% in localization accuracy" is not easy to locate in the tables (Table 1 shows a 10.0-point IoU gap and Table 3 a 12.87-point gap). Please state the exact comparison underlying this claim or revise the wording to avoid ambiguity.

Circularity Check

1 steps flagged · score 4.0 of 10

BPPL's instance pseudo-labels are constructed from the same attention scores that the loss then optimizes, forming a self-training loop; the final localization claim is empirically evaluated against real labels, so the circularity is partial rather than total.

  1. other [Section 3.3, Eq. (7)-(8)]
    "Then, we define r = xbag − ¯xbag and ˜r = ˜xbag − ¯xbag. With these features, we compute the binary class probability for xm as: pm = S([sim(xm − ¯x, r)T , sim(xm − ¯x, ˜r)T ]), (7) ... Then, ˆym = arg max pm is utilized as the pseudo-label. ... Lm = 1 ( ˆY = Y )1 (ˆpm > τ1)H(ˆym, αm), (8)"

    The pseudo-labels y_hat_m in Eq. (7) are computed from bag prototypes xbag and x_tilde_bag, which are attention-weighted combinations of the instances (xbag = sum a_m x_m; Eq. (5) defines x_tilde_bag using reversed attention from the same normalized attention alpha). The loss in Eq. (8) is a cross-entropy between this pseudo-label and the normalized attention alpha_m, so the objective pushes the model's attention toward a target that is itself a deterministic function of that same attention. No independent instance-level supervision enters; the instance signal is a self-confirmation loop that can sharpen or entrench the model's existing attention pattern. The paper itself acknowledges in Sec.

full rationale

The only step approaching circularity is Bag Prototypes-based Pseudo-Labeling (BPPL): the instance pseudo-labels are generated from the model's own attention (via attention-weighted prototypes) and then used to train that same attention through Eq. (8). This is a self-training loop rather than a derivation from independent instance labels, and the paper's own Sec. S5 explicitly states that BPPL 'heavily relies on the accuracy of attention scores as pseudo-labels.' However, the central empirical claim is not forced: IoU, Dice, and ACCinst are evaluated against ground-truth tumor annotations, and the reported gains over the GDAT-only baseline show that the loop can fail (e.g., without Lsep or filtering, performance degrades substantially). Thus the localization improvement has independent empirical content, even though the pseudo-label supervision is self-referential. I found no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via citation; GDAT is an efficiency architecture and OWLoRA is a standard orthogonal low-rank adaptation with SVD initialization, neither of which reduces to its own output. The overall circularity score is therefore moderate: one genuinely self-referential training mechanism, but with independent evaluation and a candid limitation statement.

Assumptions & free parameters 9 free parameters · 6 assumptions · 0 invented entities

The central claim rests on assumptions about attention-as-probability, synthetic normal slides, and low-rank task subspaces, plus a large set of tuned hyperparameters. None of these is derived from first principles; they are design choices validated empirically on the same datasets used for the headline results.

free parameters (9)
  • eta (GDAT residual weight) = 0.1
    Eq. (4); ablation in Tab. S5 selects 0.1 on the evaluation dataset.
  • T (BPPL temperature) = 0.5
    Eq. (7); Tab. S7 shows performance is sensitive to T and selects 0.5.
  • tau (BPPL pseudo-label threshold) = 0.35
    Defines tau1 in Eq. (8); selected by ablation Tab. S7.
  • lambda1 (instance loss weight) = 0.5
    Eq. (9); selected by ablation Tab. S8.
  • lambda2 (separation loss weight) = 1.0
    Eq. (9); selected by ablation Tab. S8.
  • epsilon (SVD energy threshold) = 0.99
    OWLoRA first-task basis truncation; selected by ablation Tab. S9.
  • d (LoRA rank) = 16
    OWLoRA basis rank for subsequent tasks; selected by ablation Tab. S9.
  • lambda3 (orthogonality loss weight) = 1.0
    Eq. (10); selected by ablation Tab. S10.
  • GDAT grouping window F and number of layers = F=16, layers=2
    Tab. S6 selects F=16 and 2 layers; this affects attention complexity and localization quality.
assumptions (6)
  • domain assumption MIL attention scores can be treated as instance class probabilities after normalization.
    Sec. 3.3 assumes a well-localizing model so that high alpha_m marks positive instances; this is the precondition for BPPL.
  • domain assumption Bag prototypes formed by attention-weighted averaging separate positive and negative instances sufficiently after applying Lsep.
    Eqs. (5)-(7) rely on x_bag and x_tilde_bag being class-representative despite attention noise.
  • domain assumption The constructed PAIP normal slides, made by removing tumor regions from tumor slides, are a valid proxy for real normal slides.
    Supp. S1.1 describes the random split and deletion of tumor pixels to create normal bags; the evaluation uses this constructed distribution.
  • domain assumption Each sequential task's weight change is low-rank and can be captured by d-dimensional orthogonal bases.
    OWLoRA uses SVD truncation and rank-d updates; Supp. S2.3 argues low-rankedness from singular value plots.
  • domain assumption Orthogonalizing new-task bases with respect to old bases reduces forgetting without blocking adaptation.
    Eqs. (10)-(11) enforce intra- and inter-orthogonality; this is a geometric assumption about task subspaces.
  • standard math Standard SVD and gradient projection properties hold.
    Used for W = USV^T and gradient projection in Eq. (11).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Continual Multiple Instance Learning with Enhanced Localization for Histopathological Whole Slide Image Analysis." pith.science (2026). https://pith.science/paper/PKI33Q3F

@misc{pith2026250702395,
  author       = {Pith},
  title        = {Pith review of: Continual Multiple Instance Learning with Enhanced Localization for Histopathological Whole Slide Image Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PKI33Q3F}},
  note         = {Machine review of arXiv:2507.02395}
}
abstract

Multiple instance learning (MIL) significantly reduced annotation costs via bag-level weak labels for large-scale images, such as histopathological whole slide images (WSIs). However, its adaptability to continual tasks with minimal forgetting has been rarely explored, especially on instance classification for localization. Weakly incremental learning for semantic segmentation has been studied for continual localization, but it focused on natural images, leveraging global relationships among hundreds of small patches (e.g., $16 \times 16$) using pre-trained models. This approach seems infeasible for MIL localization due to enormous amounts ($\sim 10^5$) of large patches (e.g., $256 \times 256$) and no available global relationships such as cancer cells. To address these challenges, we propose Continual Multiple Instance Learning with Enhanced Localization (CoMEL), an MIL framework for both localization and adaptability with minimal forgetting. CoMEL consists of (1) Grouped Double Attention Transformer (GDAT) for efficient instance encoding, (2) Bag Prototypes-based Pseudo-Labeling (BPPL) for reliable instance pseudo-labeling, and (3) Orthogonal Weighted Low-Rank Adaptation (OWLoRA) to mitigate forgetting in both bag and instance classification. Extensive experiments on three public WSI datasets demonstrate superior performance of CoMEL, outperforming the prior arts by up to $11.00\%$ in bag-level accuracy and up to $23.4\%$ in localization accuracy under the continual MIL setup.

Figures

Figures reproduced from arXiv: 2507.02395 by the authors.

Figure 1
Figure 1. We propose Continual Multiple instance learning with [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of Continual Multiple instance learning with [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Components of (a) Grouped Double Attention Transformer (GDAT) and (b) Bag Prototypes-based Pseudo-Labeling (BPPL) for [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustration of Orthogonal Weighted Low-Rank Adapta [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Instance-level classification accuracy (ACC [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results of localization across sequential organ datasets under continual MIL setup. Each column shows localization [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

78 extracted references · 77 canonical work pages

  1. [1]

    Memory aware synapses: Learning what (not) to forget

    Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, and Tinne Tuytelaars. Memory aware synapses: Learning what (not) to forget. ECCV, 2018. 1, 2

    work page 2018

  2. [2]

    Continual learning strategies for cancer- independent detection of lymph node metastases

    P ´eter B ´andi, Maschenka Balkenhol, Marcory Van Dijk, Michel Kok, Bram van Ginneken, Jeroen van der Laak, and Geert Litjens. Continual learning strategies for cancer- independent detection of lymph node metastases. Medical Image Analysis, 85:102755, 2023. 2

    work page 2023

  3. [3]

    Diagnos- tic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer

    Babak Ehteshami Bejnordi, Mitko Veta, Paul Johannes Van Diest, Bram Van Ginneken, Nico Karssemeijer, Geert Litjens, Jeroen AWM Van Der Laak, Meyke Hermsen, Quirine F Manson, Maschenka Balkenhol, et al. Diagnos- tic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer. Jama, 318(22):2199–2210, 2017. 2, 6, 12

    work page 2017

  4. [4]

    Fal- former: Feature-aware landmarks self-attention for whole- slide image classification

    Doanh C Bui, Trinh Thi Le Vuong, and Jin Tae Kwak. Fal- former: Feature-aware landmarks self-attention for whole- slide image classification. MICCAI, 2024. 3, 7

    work page 2024

  5. [5]

    Dark experience for general continual learning: a strong, simple baseline

    Pietro Buzzega, Matteo Boschini, Angelo Porrello, Davide Abati, and Simone Calderara. Dark experience for general continual learning: a strong, simple baseline. NeurIPS, 33,

    work page

  6. [6]

    New insights on reducing abrupt representation change in online continual learning

    Lucas Caccia, Rahaf Aljundi, Nader Asadi, Tinne Tuyte- laars, Joelle Pineau, and Eugene Belilovsky. New insights on reducing abrupt representation change in online continual learning. In ICLR, 2022. 6

    work page 2022

  7. [7]

    Sm: en- hanced localization in multiple instance learning for medical imaging classification

    Francisco M Castro-Mac ´ıas, Pablo Morales-Alvarez, Yunan Wu, Rafael Molina, and Aggelos Katsaggelos. Sm: en- hanced localization in multiple instance learning for medical imaging classification. NeurIPS, 2024. 1, 2, 8, 16

    work page 2024

  8. [8]

    Incremental learning in semantic segmentation from image labels

    Fabio Cermelli, Dario Fontanel, Antonio Tavera, Marco Cic- cone, and Barbara Caputo. Incremental learning in semantic segmentation from image labels. In CVPR, 2022. 1

    work page 2022

Show all 78 references
  1. [9]

    Riemannian walk for incremen- tal learning: Understanding forgetting and intransigence

    Arslan Chaudhry, Puneet K Dokania, Thalaiyasingam Ajan- than, and Philip HS Torr. Riemannian walk for incremen- tal learning: Understanding forgetting and intransigence. ECCV, 2018. 2

    2018

  2. [10]

    Efficient lifelong learning with a- gem

    Arslan Chaudhry, Marc’Aurelio Ranzato, Marcus Rohrbach, and Mohamed Elhoseiny. Efficient lifelong learning with a- gem. ICLR, 2018. 6, 7

    2018

  3. [11]

    Using hindsight to anchor past knowledge in continual learning

    Arslan Chaudhry, Albert Gordo, Puneet Dokania, Philip Torr, and David Lopez-Paz. Using hindsight to anchor past knowledge in continual learning. AAAI, 35, 2021. 1

    2021

  4. [12]

    Scaling vision transformers to gigapixel images via hierarchical self-supervised learning

    Richard J Chen, Chengkuan Chen, Yicong Li, Tiffany Y Chen, Andrew D Trister, Rahul G Krishnan, and Faisal Mah- mood. Scaling vision transformers to gigapixel images via hierarchical self-supervised learning. CVPR, 2022. 1, 2, 3, 4, 12

    2022

  5. [13]

    Towards a general-purpose foundation model for computational pathology

    Richard J Chen, Tong Ding, Ming Y Lu, Drew FK Williamson, Guillaume Jaume, Andrew H Song, Bowen Chen, Andrew Zhang, Daniel Shao, Muhammad Shaban, et al. Towards a general-purpose foundation model for computational pathology. Nature Medicine, 30(3):850–862,

  6. [14]

    cdp-mil: Robust multiple instance learning via cascaded dirichlet process

    Yihang Chen, Tsai Hor Chan, Guosheng Yin, Yuming Jiang, and Lequan Yu. cdp-mil: Robust multiple instance learning via cascaded dirichlet process. ECCV, 2024. 2

    2024

  7. [15]

    A continual learning survey: Defying for- getting in classification tasks

    Matthias De Lange, Rahaf Aljundi, Marc Masana, Sarah Parisot, Xu Jia, Ale ˇs Leonardis, Gregory Slabaugh, and Tinne Tuytelaars. A continual learning survey: Defying for- getting in classification tasks. Transactions on Pattern Anal- ysis and Machine Intelligence, 44(7):3366–33...

    2021

  8. [16]

    Deep miml network

    Ji Feng and Zhi-Hua Zhou. Deep miml network. AAAI, 2017. 2

    2017

  9. [17]

    Queryable prototype multiple instance learning with vision-language models for incremental whole slide image classification

    Jiaxiang Gou, Luping Ji, Pei Liu, and Mao Ye. Queryable prototype multiple instance learning with vision-language models for incremental whole slide image classification. arXiv preprint arXiv:2410.10573, 2024. 1, 2, 5, 6, 16

    2024 arXiv

  10. [18]

    Agent attention: On the integration of softmax and linear attention

    Dongchen Han, Tianzhu Ye, Yizeng Han, Zhuofan Xia, Siyuan Pan, Pengfei Wan, Shiji Song, and Gao Huang. Agent attention: On the integration of softmax and linear attention. ECCV, 2024. 3

    2024

  11. [19]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. CVPR, 2016. 8

    2016

  12. [20]

    Class- incremental continual learning for instance segmentation with image-level weak supervision

    Yu-Hsing Hsieh, Guan-Sheng Chen, Shun-Xian Cai, Ting- Yun Wei, Huei-Fang Yang, and Chu-Song Chen. Class- incremental continual learning for instance segmentation with image-level weak supervision. In ICCV, 2023. 1

    2023

  13. [21]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. ICLR, 2022. 2, 5, 6, 7

    2022

  14. [22]

    Conslide: Asynchronous hier- archical interaction transformer with breakup-reorganize re- hearsal for continual whole slide image analysis

    Yanyan Huang, Weiqin Zhao, Shujun Wang, Yu Fu, Yum- ing Jiang, and Lequan Yu. Conslide: Asynchronous hier- archical interaction transformer with breakup-reorganize re- hearsal for continual whole slide image analysis. ICCV,

  15. [23]

    A visual-language foundation model for pathology image analysis using medical twitter

    Zhi Huang, Federico Bianchi, Mert Yuksekgonul, Thomas J Montine, and James Zou. A visual-language foundation model for pathology image analysis using medical twitter. Nature Medicine, 29(9):2307–2316, 2023. 8 9

    2023

  16. [24]

    Attention-based deep multiple instance learning

    Maximilian Ilse, Jakub Tomczak, and Max Welling. Attention-based deep multiple instance learning. ICML,

  17. [25]

    Additive mil: Intrinsically interpretable multiple instance learning for pathology

    Syed Ashar Javed, Dinkar Juyal, Harshith Padigela, Amaro Taylor-Weiner, Limin Yu, and Aaditya Prakash. Additive mil: Intrinsically interpretable multiple instance learning for pathology. NeurIPS, 2022. 2

    2022

  18. [26]

    Vi- sual prompt tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Vi- sual prompt tuning. In European conference on computer vision, pages 709–727. Springer, 2022. 2

    2022

  19. [27]

    Paip 2020: Microsatellite instability prediction in colorectal cancer

    Kyungmo Kim, Kyoungbun Lee, Sungduk Cho, Dong Un Kang, Seongkeun Park, Yunsook Kang, Hyunjeong Kim, Gheeyoung Choe, Kyung Chul Moon, Kyu Sang Lee, et al. Paip 2020: Microsatellite instability prediction in colorectal cancer. Medical Image Analysis, 89:102886, 2023. 6, 12

    2020

  20. [28]

    Paip 2019: Liver cancer segmentation chal- lenge

    Yoo Jung Kim, Hyungjoon Jang, Kyoungbun Lee, Seongkeun Park, Sung-Gyu Min, Choyeon Hong, Jeong Hwan Park, Kanggeun Lee, Jisoo Kim, Wonjae Hong, et al. Paip 2019: Liver cancer segmentation chal- lenge. Medical image analysis , 67:101854, 2021. 2, 6, 12

    2019

  21. [29]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. ICLR, 2015. 13

    2015

  22. [30]

    Overcoming catastrophic forgetting in neu- ral networks

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska- Barwinska, et al. Overcoming catastrophic forgetting in neu- ral networks. PNAS, 2017. 2, 6

    2017

  23. [31]

    Online continual learning on class incremental blurry task configuration with anytime inference

    Hyunseo Koh, Dahyun Kim, Jung-Woo Ha, and Jonghyun Choi. Online continual learning on class incremental blurry task configuration with anytime inference. In ICLR, 2022. 17

    2022

  24. [32]

    Mixture of experts meets prompt-based continual learning

    Minh Le, Huy Nguyen, Thien Trang Nguyen Vu, Huyen Trang Pham, Linh Ngo Van, Nhat Ho, et al. Mixture of experts meets prompt-based continual learning. NeurIPS,

  25. [33]

    Online continual learning on hierarchical label expansion

    Byung Hyun Lee, Okchul Jung, Jonghyun Choi, and Se Young Chun. Online continual learning on hierarchical label expansion. In Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision, pages 11761–11770,

  26. [34]

    Dou- bly perturbed task free continual learning

    Byung Hyun Lee, Min-hwan Oh, and Se Young Chun. Dou- bly perturbed task free continual learning. In AAAI, 2024. 1, 2

    2024

  27. [35]

    Dual-stream multiple instance learning network for whole slide image classifica- tion with self-supervised contrastive learning

    Bin Li, Yin Li, and Kevin W Eliceiri. Dual-stream multiple instance learning network for whole slide image classifica- tion with self-supervised contrastive learning. CVPR, 2021. 1, 2, 3, 4, 8, 13, 16

    2021

  28. [36]

    Dt-mil: deformable transformer for multi-instance learning on histopathological image

    Hang Li, Fan Yang, Yu Zhao, Xiaohan Xing, Jun Zhang, Mingxuan Gao, Junzhou Huang, Liansheng Wang, and Jian- hua Yao. Dt-mil: deformable transformer for multi-instance learning on histopathological image. MICCAI, 2021. 1, 2

    2021

  29. [37]

    Task-specific fine-tuning via variational information bottle- neck for weakly-supervised pathology whole slide image classification

    Honglin Li, Chenglu Zhu, Yunlong Zhang, Yuxuan Sun, Zhongyi Shui, Wenwei Kuang, Sunyi Zheng, and Lin Yang. Task-specific fine-tuning via variational information bottle- neck for weakly-supervised pathology whole slide image classification. CVPR, pages 7454–7463, 2023. 2

    2023

  30. [38]

    Learning without forgetting

    Zhizhong Li and Derek Hoiem. Learning without forgetting. Transactions on Pattern Analysis and Machine Intelligence, 40(12), 2017. 2, 6, 7

    2017

  31. [39]

    Inflora: Interference-free low-rank adaptation for continual learning

    Yan-Shuo Liang and Wu-Jun Li. Inflora: Interference-free low-rank adaptation for continual learning. CVPR, 2024. 2, 5, 6, 7

    2024

  32. [40]

    Learning from the web: Language drives weakly- supervised incremental learning for semantic segmentation

    Chang Liu, Giulia Rizzoli, Pietro Zanuttigh, Fu Li, and Yi Niu. Learning from the web: Language drives weakly- supervised incremental learning for semantic segmentation. In ECCV, 2024. 1

    2024

  33. [41]

    Multiple instance learning via iterative self-paced supervised contrastive learning

    Kangning Liu, Weicheng Zhu, Yiqiu Shen, Sheng Liu, Narges Razavian, Krzysztof J Geras, and Carlos Fernandez- Granda. Multiple instance learning via iterative self-paced supervised contrastive learning. CVPR, 2023. 1, 2

    2023

  34. [42]

    Weakly-supervised residual evidential learning for multi-instance uncertainty estimation

    Pei Liu and Luping Ji. Weakly-supervised residual evidential learning for multi-instance uncertainty estimation. ICML,

  35. [43]

    Advmil: Adversarial multiple instance learning for the survival analysis on whole- slide images

    Pei Liu, Luping Ji, Feng Ye, and Bo Fu. Advmil: Adversarial multiple instance learning for the survival analysis on whole- slide images. Medical Image Analysis, 91:103020, 2024. 1

    2024

  36. [44]

    Gradient episodic memory for continual learning

    David Lopez-Paz and Marc’Aurelio Ranzato. Gradient episodic memory for continual learning. NeurIPS, 2017. 2

    2017

  37. [45]

    Data-efficient and weakly supervised computational pathology on whole- slide images

    Ming Y Lu, Drew FK Williamson, Tiffany Y Chen, Richard J Chen, Matteo Barbieri, and Faisal Mahmood. Data-efficient and weakly supervised computational pathology on whole- slide images. Nature biomedical engineering, 5(6):555–570,

  38. [46]

    A visual- language foundation model for computational pathology

    Ming Y Lu, Bowen Chen, Drew FK Williamson, Richard J Chen, Ivy Liang, Tong Ding, Guillaume Jaume, Igor Odintsov, Long Phi Le, Georg Gerber, et al. A visual- language foundation model for computational pathology. Nature Medicine, 30(3):863–874, 2024. 5, 8

    2024

  39. [47]

    Packnet: Adding mul- tiple tasks to a single network by iterative pruning

    Arun Mallya and Svetlana Lazebnik. Packnet: Adding mul- tiple tasks to a single network by iterative pruning. CVPR,

  40. [48]

    Piggy- back: Adapting a single network to multiple tasks by learn- ing to mask weights

    Arun Mallya, Dillon Davis, and Svetlana Lazebnik. Piggy- back: Adapting a single network to multiple tasks by learn- ing to mask weights. ECCV, 2018. 2

    2018

  41. [49]

    Learn more, but bother less: parameter efficient continual learning

    Fuli Qiao and Mehrdad Mahdavi. Learn more, but bother less: parameter efficient continual learning. NeurIPS, 2024. 1, 2, 5

    2024

  42. [50]

    Bi- directional weakly supervised knowledge distillation for whole slide image classification

    Linhao Qu, Manning Wang, Zhijian Song, et al. Bi- directional weakly supervised knowledge distillation for whole slide image classification. NeurIPS, 2022. 2

    2022

  43. [51]

    Experience replay for continual learning

    David Rolnick, Arun Ahuja, Jonathan Schwarz, Timothy Lil- licrap, and Gregory Wayne. Experience replay for continual learning. NeurIPS, 2019. 2, 6, 7

    2019

  44. [52]

    Overcoming catastrophic forgetting with hard attention to the task

    Joan Serra, Didac Suris, Marius Miron, and Alexandros Karatzoglou. Overcoming catastrophic forgetting with hard attention to the task. ICLR, 2018. 2, 7

    2018

  45. [53]

    Transmil: Transformer based correlated multiple instance learning for whole slide image classification

    Zhuchen Shao, Hao Bian, Yang Chen, Yifeng Wang, Jian Zhang, Xiangyang Ji, et al. Transmil: Transformer based correlated multiple instance learning for whole slide image classification. NeurIPS, 2021. 2, 3, 4, 8, 16

    2021

  46. [54]

    Loss-based attention for deep multi- ple instance learning

    Xiaoshuang Shi, Fuyong Xing, Yuanpu Xie, Zizhao Zhang, Lei Cui, and Lin Yang. Loss-based attention for deep multi- ple instance learning. AAAI, 2020. 2 10

    2020

  47. [55]

    Tendency-driven mutual exclusivity for weakly supervised incremental semantic segmentation

    Chongjie Si, Xuehui Wang, Xiaokang Yang, and Wei Shen. Tendency-driven mutual exclusivity for weakly supervised incremental semantic segmentation. In ECCV, 2024. 1

    2024

  48. [56]

    Coda-prompt: Contin- ual decomposed attention-based prompting for rehearsal-free continual learning

    James Seale Smith, Leonid Karlinsky, Vyshnavi Gutta, Paola Cascante-Bonilla, Donghyun Kim, Assaf Arbelle, Rameswar Panda, Rogerio Feris, and Zsolt Kira. Coda-prompt: Contin- ual decomposed attention-based prompting for rehearsal-free continual learning. ICCV, 2023. 2

    2023

  49. [57]

    Multiple instance learning framework with masked hard instance mining for whole slide image classification

    Wenhao Tang, Sheng Huang, Xiaoxian Zhang, Fengtao Zhou, Yi Zhang, and Bo Liu. Multiple instance learning framework with masked hard instance mining for whole slide image classification. ICCV, 2023. 2, 3, 13

    2023

  50. [58]

    Feature re-embedding: Towards foun- dation model-level performance in computational pathology

    Wenhao Tang, Fengtao Zhou, Sheng Huang, Xiang Zhu, Yi Zhang, and Bo Liu. Feature re-embedding: Towards foun- dation model-level performance in computational pathology. CVPR, 2024. 1, 3, 4, 8, 16

    2024

  51. [59]

    Three continual learning scenarios

    Gido M van de Ven and Andreas S Tolias. Three continual learning scenarios. NeurIPS Continual Learning Workshop, 1(9), 2018. 2

    2018

  52. [60]

    Rethinking multiple instance learning for whole slide image classification: A bag-level classifier is a good instance-level teacher

    Hongyi Wang, Luyang Luo, Fang Wang, Ruofeng Tong, Yen-Wei Chen, Hongjie Hu, Lanfen Lin, and Hao Chen. Rethinking multiple instance learning for whole slide image classification: A bag-level classifier is a good instance-level teacher. Transactions on Medical Imaging, 2024. 1, 2

    2024

  53. [61]

    Afec: Active forgetting of negative transfer in continual learning

    Liyuan Wang, Mingtian Zhang, Zhongfan Jia, Qian Li, Chenglong Bao, Kaisheng Ma, Jun Zhu, and Yi Zhong. Afec: Active forgetting of negative transfer in continual learning. NeurIPS, 2021. 2

    2021

  54. [62]

    Hierarchical decomposition of prompt-based continual learning: Rethinking obscured sub- optimality

    Liyuan Wang, Jingyi Xie, Xingxing Zhang, Mingyi Huang, Hang Su, and Jun Zhu. Hierarchical decomposition of prompt-based continual learning: Rethinking obscured sub- optimality. NeurIPS, 2023. 2

    2023

  55. [63]

    Attriclip: A non-incremental learner for incremen- tal knowledge learning

    Runqi Wang, Xiaoyue Duan, Guoliang Kang, Jianzhuang Liu, Shaohui Lin, Songcen Xu, Jinhu L ¨u, and Baochang Zhang. Attriclip: A non-incremental learner for incremen- tal knowledge learning. CVPR, 2023. 2

    2023

  56. [64]

    Or- thogonal subspace learning for language model continual learning

    Xiao Wang, Tianze Chen, Qiming Ge, Han Xia, Rong Bao, Rui Zheng, Qi Zhang, Tao Gui, and Xuanjing Huang. Or- thogonal subspace learning for language model continual learning. EMNLP, 2023. 2

    2023

  57. [65]

    Dualprompt: Complementary prompting for rehearsal-free continual learning

    Zifeng Wang, Zizhao Zhang, Sayna Ebrahimi, Ruoxi Sun, Han Zhang, Chen-Yu Lee, Xiaoqi Ren, Guolong Su, Vin- cent Perot, Jennifer Dy, et al. Dualprompt: Complementary prompting for rehearsal-free continual learning. In ECCV,

  58. [66]

    Learning to prompt for continual learning

    Zifeng Wang, Zizhao Zhang, Chen-Yu Lee, Han Zhang, Ruoxi Sun, Xiaoqi Ren, Guolong Su, Vincent Perot, Jen- nifer Dy, and Tomas Pfister. Learning to prompt for continual learning. In CVPR, 2022. 2

    2022

  59. [67]

    The cancer genome atlas pan-cancer analysis project

    John N Weinstein, Eric A Collisson, Gordon B Mills, Kenna R Shaw, Brad A Ozenberger, Kyle Ellrott, Ilya Shmulevich, Chris Sander, and Joshua M Stuart. The cancer genome atlas pan-cancer analysis project. Nature genetics, 45(10):1113–1120, 2013. 2, 6, 12

    2013

  60. [68]

    Wisepaip: Whole-slide image standard evalua- tion for pathology ai platform

    WISEPAIP. Wisepaip: Whole-slide image standard evalua- tion for pathology ai platform. http://www.wisepaip.org/. 2, 6, 12

  61. [69]

    Nystr¨omformer: A nystr¨om-based algorithm for approximat- ing self-attention

    Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, and Vikas Singh. Nystr¨omformer: A nystr¨om-based algorithm for approximat- ing self-attention. AAAI, 2021. 3

    2021

  62. [70]

    Rcs-prompt: Learning prompt to rearrange class space for prompt-based continual learning

    Longrong Yang, Hanbin Zhao, Yunlong Yu, Xiaodong Zeng, and Xi Li. Rcs-prompt: Learning prompt to rearrange class space for prompt-based continual learning. ECCV, 2024. 2

    2024

  63. [71]

    Foundation model drives weakly incremental learning for semantic segmentation

    Chaohui Yu, Qiang Zhou, Jingliang Li, Jianlong Yuan, Zhibin Wang, and Fan Wang. Foundation model drives weakly incremental learning for semantic segmentation. In CVPR, 2023. 1

    2023

  64. [72]

    Bayes- mil: A new probabilistic perspective on attention-based mul- tiple instance learning for whole slide images

    Cui Yufei, Ziquan Liu, Xiangyu Liu, Xue Liu, Cong Wang, Tei-Wei Kuo, Chun Jason Xue, and Antoni B Chan. Bayes- mil: A new probabilistic perspective on attention-based mul- tiple instance learning for whole slide images. ICLR, 2022. 2

    2022

  65. [73]

    Contin- ual learning through synaptic intelligence

    Friedemann Zenke, Ben Poole, and Surya Ganguli. Contin- ual learning through synaptic intelligence. ICML, 2017. 2

    2017

  66. [74]

    Dtfd- mil: Double-tier feature distillation multiple instance learn- ing for histopathology whole slide image classification

    Hongrun Zhang, Yanda Meng, Yitian Zhao, Yihong Qiao, Xiaoyun Yang, Sarah E Coupland, and Yalin Zheng. Dtfd- mil: Double-tier feature distillation multiple instance learn- ing for histopathology whole slide image classification. CVPR, 2022. 1, 2, 13

    2022

  67. [75]

    Setmil: spa- tial encoding transformer-based multiple instance learning for pathological image analysis

    Yu Zhao, Zhenyu Lin, Kai Sun, Yidan Zhang, Junzhou Huang, Liansheng Wang, and Jianhua Yao. Setmil: spa- tial encoding transformer-based multiple instance learning for pathological image analysis. MICCAI, 2022. 2

    2022

  68. [76]

    A graph-transformer for whole slide image classifica- tion

    Yi Zheng, Rushin H Gindra, Emily J Green, Eric J Burks, Margrit Betke, Jennifer E Beane, and Vijaya B Kolacha- lama. A graph-transformer for whole slide image classifica- tion. Transactions on Medical Imaging, 41(11):3003–3015,

  69. [77]

    Deep multi-instance networks with sparse label assignment for whole mammogram classification

    Wentao Zhu, Qi Lou, Yeeleng Scott Vang, and Xiaohui Xie. Deep multi-instance networks with sparse label assignment for whole mammogram classification. MICCAI, 2017. 2

    2017

  70. [78]

    Dgr-mil: Exploring diverse global representation in multiple instance learning for whole slide image classification

    Wenhui Zhu, Xiwen Chen, Peijie Qiu, Aristeidis Sotiras, Abolfazl Razi, and Yalin Wang. Dgr-mil: Exploring diverse global representation in multiple instance learning for whole slide image classification. ECCV, 2024. 1 11 Continual Multiple Instance Learning with Enhanced Local...

    2024

Pith tools

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