Pith. sign in

REVIEW 4 major objections 3 minor 58 references

Multiple quantum spin Hall states and topological current divider in Twisted Bilayer WSe$_2$

T0 review · 4 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read Twisted bilayer WSe₂ hosts double and quadruple quantum spin Hall states, carried by moiré edge states that survive non-magnetic disorder.

desk verdict The abstract promises real advances in twisted TMD QSH physics, but the supplied full text is an unrelated computer-vision paper, so the claims are currently unverifiable. read the letter →

arxiv 2508.05092 v1 pith:PZMHI3GM submitted 2025-08-07 cond-mat.mes-hall

classification cond-mat.mes-hall PACS 73.43.-f72.25.-b
keywords twistedbilayerWSe2quantumspinHallmoireedgestateshelicalchannelstopologicalcurrentdividertransitionmetaldichalcogenidessuperlatticedissipationlessspintronics
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 argues that twisted bilayer WSe₂, already known to host a single quantum spin Hall (QSH) state, can be driven into “double” and “quartuple” QSH states with two or four pairs of counter-propagating helical edge channels. Its central claim is that these channels are not pinned to the physical sample edge: charge carriers flow along the high-potential points of the moiré superlattice boundary, undergoing interlayer transitions as they propagate, which the authors call moiré edge states. According to the calculation, these states survive non-magnetic disorder, and the double QSH state is more robust than the single. At a twist angle of 2.45°, a surface gate switches between single and double QSH states, and a five-terminal device built on this switching acts as a topological current divider for dissipationless spintronics.

What carries the argument

The moiré superlattice of the twisted WSe₂ bilayer is the load-bearing structure: its periodic potential reshapes the effective edge so that helical channels run along the high-potential points of the superlattice boundary rather than the physical edge. The multi-pair hierarchy—single, double, quartuple—corresponds to the number of counter-propagating helical channel pairs (1, 2, 4) in the QSH state, and the gate-controlled transition at 2.45° is what enables the proposed divider device.

What would settle it

Measure the two-terminal conductance of a twisted WSe₂ bilayer at 2.45° across a gate sweep. The claim predicts a $4e^2/h$ plateau (double QSH) alongside the $2e^2/h$ single-QSH plateau, and an $8e^2/h$ plateau at the quartuple setting; observing only $2e^2/h$, or a $4e^2/h$ plateau that a gate cannot remove, would refute it. Local imaging (STM or scanning-gate) that places the channels at the physical edge rather than at the moiré potential maxima would likewise falsify the moiré-edge-state picture.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is a hierarchy of helical edge states in twisted bilayer WSe₂: conventional single QSH (one counter-propagating helical pair), emergent double QSH (two pairs), and quartuple QSH (four pairs). The carriers in these states sit at the high-potential point of the moiré superlattice boundary, not at the physical edge of the sample, and propagate forward by switching layers. The states remain conducting in the presence of non-magnetic disorder, with the double QSH state more robust than the single. At the 2.45° twist angle, a gate on the surface toggles between the single and double QSH states, which the authors use to propose a five-terminal device operatin

Load-bearing premise

Everything follows from the model of the twisted WSe₂ moiré pattern: if the Hamiltonian, moiré potential profile, or disorder treatment used in the calculation does not match what real twisted WSe₂ bilayers do, the predicted double and quadruple channels, their location at the moiré high-potential points, and the 2.45° switching angle would not occur in experiments.

Editorial extensions

If this is right

  • A two-terminal conductance measurement should show quantized plateaus of $2e^2/h$, $4e^2/h$, and $8e^2/h$ as the twisted bilayer passes through the single, double, and quartuple QSH regimes.
  • The gate-controlled transition at 2.45° gives a working switch between one and two helical pairs, enabling a five-terminal topological current divider.
  • Because double QSH is claimed more robust than single QSH against non-magnetic disorder, devices with more helical pairs could be more forgiving of disorder, not less.
  • Moiré edge states decouple the topological channel from the physical sample boundary, so the effective circuit is defined by the superlattice rather than by lithographic edges.

Reading between the lines

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

  • If moiré edge states localize at potential maxima, then tuning the moiré potential—through twist angle, pressure, or a dielectric environment—should reconfigure the effective edge, a consequence the paper does not develop.
  • A natural test of the robustness hierarchy is to compare the conductance plateau quality (plateau width versus disorder strength) across single and double QSH devices in the same sample by electrostatic gating.
  • The supplied full text is a different manuscript (on few-shot class-incremental learning), so the abstract's WSe₂ claims are not backed by the body provided; the Hamiltonian, moiré potential profile, and disorder model must be checked in the actual paper.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 3 minor

Summary. The submission, identified as arXiv:2508.05092 (cond-mat.mes-hall), claims in its abstract that twisted bilayer WSe2 hosts, in addition to the usual single quantum spin Hall (QSH) state, double and 'quartuple' QSH states with two and four pairs of counter-propagating helical edge channels. It further claims that these channels are not at the physical edge but at high-potential points of the moire superlattice boundary ('moire edge states'), that they survive nonmagnetic disorder with double-QSH robustness exceeding single-QSH robustness, that a surface gate at twist angle 2.45° can switch between single and double QSH states, and that a five-terminal device can act as a topological current divider. However, the full text supplied with the submission is not the WSe2 paper; it is arXiv:2508.05094, a computer-vision manuscript on few-shot class-incremental learning. No Hamiltonian, moire potential, band structure, boundary construction, disorder model, transport calculation, or device simulation is present. The abstract is therefore the only reviewable content.

Significance. The physical claims are potentially significant: multiple QSH edge-channel multiplicities and a gate-controlled transition in a twisted TMD platform would be of genuine interest, and a topological current divider would be a useful conceptual device. The abstract also makes falsifiable predictions (edge-pair counts, localization at moire potential maxima, a 2.45° transition, and a robustness ordering), which is commendable in principle. However, the submission contains no derivations, numerical results, reproducible code, or parameter-free statements that could substantiate these predictions. Because the body is an unrelated paper, the significance is entirely conditional: none of the claimed phenomena can be checked, and the modeling choices that would determine whether the outputs follow from physics are undisclosed.

major comments (4)
  1. [Full text (body)] The supplied full text is arXiv:2508.05094, an unrelated cs.CV paper on few-shot class-incremental learning. This is a load-bearing omission: the WSe2 claims in the abstract are supported by no derivable content. To assess the central claims of double and quadruple QSH states with moire-edge-channel multiplicity, the manuscript must provide the twisted-bilayer Hamiltonian, moire potential profile, boundary construction, band-structure results, and the criterion used to count helical edge pairs. None of these are present.
  2. [Abstract (2.45° transition)] The abstract pins a single-to-double QSH transition at 2.45° and states that it can be induced by adjusting a surface gate. No phase diagram, gap-closing analysis, topological invariant calculation, or gate-voltage range is given. This quantitative transition point is a central predictive claim and is currently unsupported.
  3. [Abstract (disorder robustness)] The claim that double QSH states are more robust than single QSH states under nonmagnetic disorder requires a disorder model and a transport or localization calculation (e.g., conductance quantization, edge-state decay length, or disorder-averaged transmission). The submitted text contains no disorder Hamiltonian, no disorder geometry or strength, and no transport data, so this ordering cannot be evaluated.
  4. [Abstract (moire edge states)] The new entity 'moire edge states' is introduced without a definition of the 'moire superlattice boundary' or the 'high potential point' at which the channels are said to localize, and without a calculation showing that the carriers are not at the physical edge. Since this is the paper's main novelty, its absence of support is load-bearing.
minor comments (3)
  1. [Abstract] Language issues: 'quartuple' should be 'quadruple'; 'topological current devider' should be 'current divider'; 'QSH state exist' should be 'states exist'; and the phrase 'state ex' appears truncated. The intended meaning of 'high potential point of the moire superlattice boundary' is unclear.
  2. [Abstract] 'At a twisting angle of 2.45°' should be 'twist angle'; and 'adjusting the gate on the surface' needs specification (top gate, bottom gate, or electrostatic doping) and a voltage range.
  3. [Full text (metadata)] The manuscript header and the embedded arXiv identifier are inconsistent with the announced paper: the full text is arXiv:2508.05094, not arXiv:2508.05092. The correct WSe2 manuscript file appears not to have been submitted.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found; supplied full text is an unrelated cs.CV paper, so the WSe2 derivation cannot be analyzed.

full rationale

The claimed paper is arXiv:2508.05092 (cond-mat.mes-hall) on twisted bilayer WSe2, but the supplied full text is arXiv:2508.05094v1, a computer-vision paper on few-shot class-incremental learning. No equations, Hamiltonian, moire potential, band-structure calculation, disorder model, or transport setup from the WSe2 paper are present. The abstract alone asserts the existence of double and quartuple quantum spin Hall states, moire edge states, a 2.45° gate-tunable transition, and a robustness ordering, but it provides no derivation chain whose steps could be checked for circularity. Under the rule that circularity can only be claimed when one can quote the paper and exhibit a specific reduction (e.g., an equation identical to an input by construction, or a fitted parameter renamed as a prediction), no such step can be identified. The absence of methodology is a completeness/verifiability problem, not evidence of circularity. Therefore the honest finding is no significant circularity, score 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

This ledger is necessarily provisional because only the abstract was reviewable: the Hamiltonian, moire potential construction, disorder model, and transport method are undisclosed. The listed parameters are the ones any such calculation must contain; the listed axioms are the background assumptions the claims visibly depend on. No invented physical entity with independent falsifiable evidence is provided in the abstract; the 'moire edge state' concept is the paper's own label for its computed channels, so it is inventoried with independent_evidence = False.

free parameters (4)
  • Model Hamiltonian parameters for twisted bilayer WSe2 (interlayer tunneling, onsite potentials, moire potential profile)
    The abstract does not state the Hamiltonian. Any calculation of QSH gaps in twisted TMDs depends on such parameters, and the channel multiplicity (one vs two vs four pairs) is sensitive to them.
  • Twist angle for the single-to-double transition = 2.45 degrees
    Presented as the angle at which a gate drives the transition; without a derivation it may be a selected operating point from an angle scan rather than a parameter-free prediction.
  • Disorder strength and geometry for the nonmagnetic disorder robustness claim
    The ordering 'double QSH more robust than single' requires a specific disorder model, concentration, and system size that the abstract does not disclose.
  • Gate voltage range for the transition
    The gate is the control knob for the single-to-double transition and the current divider, but its value and the electrostatics model are not given.
assumptions (4)
  • standard math Bulk-boundary correspondence counts helical edge pairs via spin Chern numbers computed from the moire band structure.
    Invoked implicitly by labeling states by the number of counter-propagating helical edge channels; requires a well-defined spin-Chern classification.
  • domain assumption A faithful moire superlattice model of twisted WSe2 at small angles (spin-orbit coupling, interlayer hybridization, possibly strain relaxation).
    All results depend on the model; the abstract gives no model details.
  • domain assumption Time-reversal symmetry and a well-defined spin projection are preserved so that the QSH classification applies.
    Necessary for helical edge channels and spin filtering; not stated in the abstract.
  • ad hoc to paper The 'moire superlattice boundary' potential used to define moire edge states is a faithful representation of the physical boundary.
    The claim that channels localize at high potential points of the superlattice boundary depends on how that boundary was constructed; if the boundary potential is imposed by the calculation, the localization could be an artifact.
invented entities (1)
  • Moire edge states
    purpose: Label for helical conducting channels located at moire potential maxima rather than at the physical edge, with interlayer hopping and continuous forward propagation.
    Introduced as a new category of edge state in the abstract. No independent falsifiable handle (such as a predicted conductance signature tied to an external observable) is given in the abstract; the only evidence cited is the authors' own calculation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multiple quantum spin Hall states and topological current divider in Twisted Bilayer WSe$_2$." pith.science (2026). https://pith.science/paper/PZMHI3GM

@misc{pith2026250805092,
  author       = {Pith},
  title        = {Pith review of: Multiple quantum spin Hall states and topological current divider in Twisted Bilayer WSe$_2$},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PZMHI3GM}},
  note         = {Machine review of arXiv:2508.05092}
}
abstract

It has been demonstrated that topological quantum spin Hall (QSH) state exist in twisted bilayers of transition metal dichalcogenides. However, a comprehensive theoretical characterization of the topological edge states remains a topic of interest and an unresolved issue. Here, the topological transport properties of the twisted WSe$_2$ bilayers are investigated. Beyond the conventional single QSH, we identify emergent double and quartuple quantum spin Hall states, hosting two and four pairs of counter-propagating helical edge channels respectively. Furthermore, the charge carriers in these edge states are not localized at edge but rather the high potential point of the moire superlattice boundary, undergoing interlayer transitions and propagating forward continuously. We term these edge states as moire edge states. These edge states can survive in non-magnetic disorder, with the robustness of double QSH states surpassing that of single QSH states. At a twisting angle of 2.45$^\circ$, the transition between the single and double QSH states can be achieved by adjusting the gate on the surface. Based on this, we propose a five-terminal device to as a topological current devider. Our findings provide support for the development of dissipationless spintronics.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

58 extracted references · 50 canonical work pages

  1. [1]

    Class-incremental learning: survey and performance evaluation on image classification,

    M. Masana, X. Liu, B. Twardowski, M. Menta, A. D. Bagdanov, and J. Van De Weijer, “Class-incremental learning: survey and performance evaluation on image classification,” IEEE Transactions on Pattern Anal- ysis and Machine Intelligence , vol. 45, no. 5, pp. 5513–5533, 2022

  2. [2]

    Class-incremental learning: A survey,

    D.-W. Zhou, Q.-W. Wang, Z.-H. Qi, H.-J. Ye, D.-C. Zhan, and Z. Liu, “Class-incremental learning: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024

  3. [3]

    Class incremental learning for light-weighted networks,

    Z. Tao, L. Yu, H. Yao, S. Huang, and C. Xu, “Class incremental learning for light-weighted networks,” IEEE Transactions on Circuits and Systems for Video Technology , 2024

  4. [4]

    Prototype augmentation and self-supervision for incremental learning,

    F. Zhu, X.-Y . Zhang, C. Wang, F. Yin, and C.-L. Liu, “Prototype augmentation and self-supervision for incremental learning,” in Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 5871–5880

  5. [5]

    Self- organizing pathway expansion for non-exemplar class-incremental learn- ing,

    K. Zhu, K. Zheng, R. Feng, D. Zhao, Y . Cao, and Z.-J. Zha, “Self- organizing pathway expansion for non-exemplar class-incremental learn- ing,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 19 204–19 213

  6. [6]

    Prototype reminiscence and augmented asymmetric knowledge aggregation for non-exemplar class-incremental learning,

    W. Shi and M. Ye, “Prototype reminiscence and augmented asymmetric knowledge aggregation for non-exemplar class-incremental learning,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 1772–1781

  7. [7]

    Reformulating classification as image-class matching for class incremental learning,

    Y . Hu, Z. Liang, X. Liu, Q. Hou, and M.-M. Cheng, “Reformulating classification as image-class matching for class incremental learning,” IEEE Transactions on Circuits and Systems for Video Technology, 2024

  8. [8]

    Non-exemplar class- incremental learning by random auxiliary classes augmentation and mixed features,

    K. Song, G. Liang, Z. Chen, and Y . Zhang, “Non-exemplar class- incremental learning by random auxiliary classes augmentation and mixed features,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 9, pp. 7830–7843, 2024

Show all 58 references
  1. [9]

    Task- adaptive saliency guidance for exemplar-free class incremental learning,

    X. Liu, J.-T. Zhai, A. D. Bagdanov, K. Li, and M.-M. Cheng, “Task- adaptive saliency guidance for exemplar-free class incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 954–23 963

  2. [10]

    Adapter merging with centroid prototype mapping for scalable class-incremental learning,

    T. Fukuda, H. Kera, and K. Kawamoto, “Adapter merging with centroid prototype mapping for scalable class-incremental learning,” in Proceed- ings of the Computer Vision and Pattern Recognition Conference , 2025, pp. 4884–4893

  3. [11]

    A survey on few- shot class-incremental learning,

    S. Tian, L. Li, W. Li, H. Ran, X. Ning, and P. Tiwari, “A survey on few- shot class-incremental learning,” Neural Networks , vol. 169, pp. 307– 324, 2024

  4. [12]

    For- ward compatible few-shot class-incremental learning,

    D.-W. Zhou, F.-Y . Wang, H.-J. Ye, L. Ma, S. Pu, and D.-C. Zhan, “For- ward compatible few-shot class-incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 9046–9056

  5. [13]

    Class-aware prompting for federated few-shot class- incremental learning,

    F.-Y . Liang, Y .-W. Zhan, J. Liu, C.-Y . Zhang, Z.-D. Chen, X. Luo, and X.-S. Xu, “Class-aware prompting for federated few-shot class- incremental learning,” IEEE Transactions on Circuits and Systems for Video Technology, 2025

  6. [14]

    Learning with fantasy: Semantic-aware virtual contrastive constraint for few-shot class- incremental learning,

    Z. Song, Y . Zhao, Y . Shi, P. Peng, L. Yuan, and Y . Tian, “Learning with fantasy: Semantic-aware virtual contrastive constraint for few-shot class- incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 24 183–24 192

  7. [15]

    Improved continually evolved clas- sifiers for few-shot class-incremental learning,

    Y . Wang, G. Zhao, and X. Qian, “Improved continually evolved clas- sifiers for few-shot class-incremental learning,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 2, pp. 1123– 1134, 2023

  8. [16]

    Few-shot class-incremental learning via training-free prototype cali- bration,

    Q.-W. Wang, D.-W. Zhou, Y .-K. Zhang, D.-C. Zhan, and H.-J. Ye, “Few-shot class-incremental learning via training-free prototype cali- bration,” Advances in Neural Information Processing Systems , vol. 36, pp. 15 060–15 076, 2023

  9. [17]

    Prompt- based concept learning for few-shot class-incremental learning,

    S. Li, F. Liu, L. Jiac, L. Li, P. Chen, X. Liu, and W. Ma, “Prompt- based concept learning for few-shot class-incremental learning,” IEEE Transactions on Circuits and Systems for Video Technology , 2025

  10. [18]

    Margin-based few-shot class- incremental learning with class-level overfitting mitigation,

    Y . Zou, S. Zhang, Y . Li, and R. Li, “Margin-based few-shot class- incremental learning with class-level overfitting mitigation,” Advances in Neural Information Processing Systems , vol. 35, pp. 27 267–27 279, 2022

  11. [19]

    Few-shot class incremental learning with attention-aware self-adaptive JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 prompt,

    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 JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 prompt,” in European Conference on Computer Vision. Springer, 2024, pp. 1–18

  12. [20]

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

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

  13. [21]

    Learning multiple layers of features from tiny images,

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

  14. [22]

    The many faces of robustness: A critical analysis of out-of-distribution generalization,

    D. Hendrycks, S. Basart, N. Mu, S. Kadavath, F. Wang, E. Dorundo, R. Desai, T. Zhu, S. Parajuli, M. Guo et al. , “The many faces of robustness: A critical analysis of out-of-distribution generalization,” in Proceedings of the IEEE/CVF International Conference on Computer Visio...

  15. [23]

    Localizing task information for improved model merging and compres- sion,

    K. Wang, N. Dimitriadis, G. Ortiz-Jimenez, F. Fleuret, and P. Frossard, “Localizing task information for improved model merging and compres- sion,” arXiv preprint arXiv:2405.07813 , 2024

  16. [24]

    Emr- merging: Tuning-free high-performance model merging,

    C. Huang, P. Ye, T. Chen, T. He, X. Yue, and W. Ouyang, “Emr- merging: Tuning-free high-performance model merging,” Advances in Neural Information Processing Systems , vol. 37, pp. 122 741–122 769, 2024

  17. [25]

    Editing models with task arithmetic,

    G. Ilharco, M. T. Ribeiro, M. Wortsman, S. Gururangan, L. Schmidt, H. Hajishirzi, and A. Farhadi, “Editing models with task arithmetic,” arXiv preprint arXiv:2212.04089 , 2022

  18. [26]

    Few-shot class-incremental learning with meta-learned class structures,

    G. Zheng and A. Zhang, “Few-shot class-incremental learning with meta-learned class structures,” in 2021 International Conference on Data Mining Workshops (ICDMW). IEEE, 2021, pp. 421–430

  19. [27]

    Incremental few-shot learning via vector quantization in deep embedded space,

    K. Chen and C.-G. Lee, “Incremental few-shot learning via vector quantization in deep embedded space,” in International Lonference on Learning Representations, 2021

  20. [28]

    Neural collapse inspired feature-classifier alignment for few-shot class incremental learn- ing,

    Y . Yang, H. Yuan, X. Li, Z. Lin, P. Torr, and D. Tao, “Neural collapse inspired feature-classifier alignment for few-shot class incremental learn- ing,” International Conference on Learning Representations , 2023

  21. [29]

    Metafscil: A meta-learning approach for few-shot class incremental learning,

    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 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 14 166–14 175

  22. [31]

    Few-shot incre- mental learning with continually evolved classifiers,

    C. Zhang, N. Song, G. Lin, Y . Zheng, P. Pan, and Y . Xu, “Few-shot incre- mental learning with continually evolved classifiers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 12 455–12 464

  23. [32]

    Dynamic support network for few-shot class incremental learning,

    B. Yang, M. Lin, Y . Zhang, B. Liu, X. Liang, R. Ji, and Q. Ye, “Dynamic support network for few-shot class incremental learning,” IEEE Trans- actions on Pattern Analysis and Machine Intelligence , vol. 45, no. 3, pp. 2945–2951, 2022

  24. [33]

    Generalized and incremental few-shot learning by explicit learning and calibration without forgetting,

    A. Kukleva, H. Kuehne, and B. Schiele, “Generalized and incremental few-shot learning by explicit learning and calibration without forgetting,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 9020–9029

  25. [34]

    Semantic-aware knowledge distillation for few-shot class- incremental learning,

    A. Cheraghian, S. Rahman, P. Fang, S. K. Roy, L. Petersson, and M. Harandi, “Semantic-aware knowledge distillation for few-shot class- incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 2534–2543

  26. [35]

    Few-shot class-incremental learning via entropy-regularized data-free replay,

    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 European Conference on Computer Vision . Springer, 2022, pp. 146– 162

  27. [36]

    Semantics- driven generative replay for few-shot class incremental learning,

    A. Agarwal, B. Banerjee, F. Cuzzolin, and S. Chaudhuri, “Semantics- driven generative replay for few-shot class incremental learning,” in Proceedings of the 30th ACM International Conference on Multimedia , 2022, pp. 5246–5254

  28. [37]

    Subspace regularizers for few-shot class incremental learning,

    D. Wijaya, A. F. Aky ¨urek, E. Akyurek, and J. Andreas, “Subspace regularizers for few-shot class incremental learning,” 2022

  29. [38]

    Warping the space: Weight space rotation for class-incremental few-shot learning,

    D.-Y . Kim, D.-J. Han, J. Seo, and J. Moon, “Warping the space: Weight space rotation for class-incremental few-shot learning,” in The Eleventh International Conference on Learning Representations , 2023

  30. [39]

    Few- shot class-incremental learning by sampling multi-phase tasks,

    D.-W. Zhou, H.-J. Ye, L. Ma, D. Xie, S. Pu, and D.-C. Zhan, “Few- shot class-incremental learning by sampling multi-phase tasks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 45, no. 11, pp. 12 816–12 831, 2022

  31. [40]

    Revisiting class-incremental learning with pre-trained models: Generalizability and adaptivity are all you need,

    D.-W. Zhou, Z.-W. Cai, H.-J. Ye, D.-C. Zhan, and Z. Liu, “Revisiting class-incremental learning with pre-trained models: Generalizability and adaptivity are all you need,” International Journal of Computer Vision , vol. 133, no. 3, pp. 1012–1032, 2025

  32. [41]

    Learning to prompt for continual learning,

    Z. Wang, Z. Zhang, C.-Y . Lee, H. Zhang, R. Sun, X. Ren, G. Su, V . Perot, J. Dy, and T. Pfister, “Learning to prompt for continual learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 139–149

  33. [42]

    Dualprompt: Complementary prompting for rehearsal-free continual learning,

    Z. Wang, Z. Zhang, S. Ebrahimi, R. Sun, H. Zhang, C.-Y . Lee, X. Ren, G. Su, V . Perot, J. Dyet al., “Dualprompt: Complementary prompting for rehearsal-free continual learning,” in European Conference on Computer Vision. Springer, 2022, pp. 631–648

  34. [43]

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

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

  35. [44]

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

    Y .-S. Liang and W.-J. Li, “Inflora: Interference-free low-rank adaptation for continual learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 638–23 647

  36. [45]

    Lora subtraction for drift-resistant space in exemplar-free continual learning,

    X. Liu and X. Chang, “Lora subtraction for drift-resistant space in exemplar-free continual learning,” in Proceedings of the Computer Vision and Pattern Recognition Conference , 2025, pp. 15 308–15 318

  37. [46]

    Sphereface: Deep hypersphere embedding for face recognition,

    W. Liu, Y . Wen, Z. Yu, M. Li, B. Raj, and L. Song, “Sphereface: Deep hypersphere embedding for face recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 212–220

  38. [47]

    Cosface: Large margin cosine loss for deep face recognition,

    H. Wang, Y . Wang, Z. Zhou, X. Ji, D. Gong, J. Zhou, Z. Li, and W. Liu, “Cosface: Large margin cosine loss for deep face recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 5265–5274

  39. [48]

    Additive margin softmax for face verification,

    F. Wang, J. Cheng, W. Liu, and H. Liu, “Additive margin softmax for face verification,” IEEE Signal Processing Letters , vol. 25, no. 7, pp. 926–930, 2018

  40. [49]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 4690–4699

  41. [50]

    Negative margin matters: Understanding margin in few-shot classification,

    B. Liu, Y . Cao, Y . Lin, Q. Li, Z. Zhang, M. Long, and H. Hu, “Negative margin matters: Understanding margin in few-shot classification,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part IV 16 . Springer, 2020, pp. 438–455

  42. [51]

    Overcoming catastrophic forgetting in neural networks,

    J. Kirkpatrick, R. Pascanu, N. Rabinowitz, J. Veness, G. Desjardins, A. A. Rusu, K. Milan, J. Quan, T. Ramalho, A. Grabska-Barwinska et al. , “Overcoming catastrophic forgetting in neural networks,” Pro- ceedings of the National Academy of Sciences , vol. 114, no. 13, pp. 3521...

  43. [52]

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

    L. Wang, J. Xie, X. Zhang, M. Huang, H. Su, and J. Zhu, “Hierarchical decomposition of prompt-based continual learning: Rethinking obscured sub-optimality,” Advances in Neural Information Processing Systems , vol. 36, pp. 69 054–69 076, 2023

  44. [53]

    Semantically- shifted incremental adapter-tuning is a continual vitransformer,

    Y . Tan, Q. Zhou, X. Xiang, K. Wang, Y . Wu, and Y . Li, “Semantically- shifted incremental adapter-tuning is a continual vitransformer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 23 252–23 262

  45. [54]

    The caltech-ucsd birds-200-2011 dataset,

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

  46. [55]

    icarl: Incremental classifier and representation learning,

    S.-A. Rebuffi, A. Kolesnikov, G. Sperl, and C. H. Lampert, “icarl: Incremental classifier and representation learning,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 2001–2010

  47. [56]

    Foster: Feature boosting and compression for class-incremental learning,

    F.-Y . Wang, D.-W. Zhou, H.-J. Ye, and D.-C. Zhan, “Foster: Feature boosting and compression for class-incremental learning,” in European Conference on Computer Vision . Springer, 2022, pp. 398–414

  48. [57]

    Adaptive decision boundary for few-shot class-incremental learning,

    L. Li, Y . Tan, S. Yang, H. Cheng, Y . Dong, and L. Yang, “Adaptive decision boundary for few-shot class-incremental learning,” in Proceed- ings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 17, 2025, pp. 18 359–18 367

  49. [58]

    Few-shot class-incremental learning,

    X. Tao, X. Hong, X. Chang, S. Dong, X. Wei, and Y . Gong, “Few-shot class-incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 12 183–12 192

  50. [59]

    Few-shot class-incremental learning from an open-set perspective,

    C. Peng, K. Zhao, T. Wang, M. Li, and B. C. Lovell, “Few-shot class-incremental learning from an open-set perspective,” in European Conference on Computer Vision . Springer, 2022, pp. 382–397

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.