Pith. sign in

REVIEW 4 major objections 8 minor 47 references

AdaptMol: Adaptive Fusion from Sequence String to Topological Structure for Few-shot Drug Discovery

T0 review · 4 major / 8 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read AdaptMol claims that adaptively fusing SMILES-sequence and molecular-graph representations through a two-level attention module gives state-of-the-art few-shot molecular property prediction on three standard benchmarks.

desk verdict The benchmark work is solid, but the paper's headline 'adaptive fusion' is a fixed scalar schedule, not input-dependent weighting; it deserves a serious referee but needs major revision. read the letter →

arxiv 2505.11878 v1 pith:Z32ZMMRD submitted 2025-05-17 cs.LG cs.AIq-bio.MN

Yifan Dai (1) , Xuanbai Ren (1) , Tengfei Ma (1) , Qipeng Yan (2) , Yiping Liu (1) , Yuansheng Liu (1) , Xiangxiang Zeng (1) ((1) College of Computer Science , Electronic Engineering
show 3 more authors
Hunan University (2) School of Biomedical Science Hunan University)
This is my paper · ORCID
classification cs.LGcs.AIq-bio.MN
keywords few-shotlearningmolecularpropertypredictionmultimodalfusionprototypicalnetworksgraphneuralSMILESattentionmechanisminterpretability
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 tries to establish that in few-shot molecular property prediction, where each new task has only a handful of labeled molecules, the limiting factor is the molecular representation itself, and that fusing two complementary views—the SMILES sequence as a global semantic summary and the molecular graph as local topological detail—through an adaptive attention module raises that ceiling. Built on a prototypical network, AdaptMol reports average ROC-AUC gains of about 4.18 percentage points over the best baseline in 5-shot settings and 2.44 percentage points in 10-shot settings across Tox21, SIDER, and MUV. If these results hold, drug-discovery teams could screen candidates against a novel endpoint with only a few assay values and still get reliable rankings, along with a chemically interpretable explanation of what the model is keying on. The paper also claims this fused representation is interpretable: a Monte Carlo tree search over bond deletions finds small active substructures that the fused model scores highly, tying the learned features to chemistry.

What carries the argument

The load-bearing component is the Adaptive Multi-level Attention (AMA) module: a two-stage attention block that first concatenates graph node embeddings weighted by one fusion weight with SMILES features weighted by another, runs multi-head self-attention with a sigmoid activation over the per-node vectors, and then applies a second, fully-connected sigmoid gate over the pooled node representation. The fusion weights in Equations (3)-(4) are meant to balance the two modalities at local and global levels, and in the implementation they are fixed constants (k=2, beta_min=0.9, beta_max=1.1) rather than learned functions of the input. AMA is what converts two encoders into a single refined molecular embedding, and the same embedding feeds both the prototype classifier and the MCTS rationale extractor.

What would settle it

A direct test is to rerun AdaptMol on the same three benchmarks with the two fusion weights both set to 1 (plain concatenation through the same attention layers) and, separately, with weights learned by a small input-dependent gating network; if the equal-weight version matches the reported ROC-AUCs, or the learned version does not improve on them, the 'adaptive' weighting is not the driver of the results.

Watch

Extended reading notes

Core claim

The paper's central claim is that the Adaptive Multi-level Attention (AMA) module—by weighting and concatenating GIN node embeddings with PCA-reduced SMILES embeddings from a large language model, then applying local- and global-level attention—produces molecular representations that a prototypical network can use to outperform existing few-shot predictors on Tox21, SIDER, and MUV under 5-shot and 10-shot supervision. The ablation study supports this by showing that removing either attention level or the adaptive weighting consistently lowers ROC-AUC, and the cross-domain experiments extend the claim to generalization across tasks from different domains. Interpretability is part of the claim: rationales extracted by MCTS from the fused model retain high predicted activity scores and highlight substructures such as amide and secondary-amine groups, which the paper reads as evidence that the fusion captures pharmacologically meaningful patterns.

Load-bearing premise

The load-bearing premise is that the hand-set fusion weights in the model, not the attention layers or the choice of encoders, are what produce the reported performance gap; if those weights were tuned on the test benchmarks, or if equal fixed weights reproduce the same scores, the adaptive-fusion claim would lose its support.

Editorial extensions

If this is right

  • On the three standard benchmarks, AdaptMol records the best ROC-AUC in every reported 5-shot and 10-shot setting, including 83.79 on Tox21 5-shot and 77.16 on MUV 10-shot.
  • The ablation study shows that removing either the local-level or the global-level attention, or the adaptive weighting, lowers ROC-AUC on every dataset, with the largest drop when the full adaptive module is replaced.
  • On a cross-domain benchmark built from classification tasks in a different domain, AdaptMol's best 10-shot ROC-AUC of 69.08 beats the two pretrained baselines it is compared with.
  • Rationales extracted by MCTS from the fused model keep high predicted activity scores and highlight chemical groups such as amide and secondary-amine moieties, which the paper presents as evidence that the fused representation tracks pharmacologically meaningful structure.

Reading between the lines

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

  • Because the fusion weights in Equations (3)-(4) are fixed constants in the implementation, the word 'adaptive' currently describes a hand-set balancing scheme, not a per-molecule decision; learning these weights from the input would test whether the mechanism itself, rather than the constants, drives the gains.
  • The paper's own limitation note (Appendix F) says the fusion is simplistic and can create redundancy for simple molecules; input-dependent fusion or learned gating is the direct next step, and the ablation setup already provides the template for evaluating it.
  • The same weighted-concatenation-plus-two-stage-attention pattern could transfer to other graph-plus-sequence domains such as proteins or materials, although AdaptMol is demonstrated only on small-molecule benchmarks.
  • Using the MCTS rationale extractor as a probe, one could compare the substructures highlighted with and without the SMILES modality on identical molecules; rationales that appear only in the fused model would quantify what the global sequence view adds beyond topology.
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

4 major / 8 minor

Summary. The paper proposes AdaptMol, a prototypical network for few-shot molecular property prediction that combines graph-based local features with SMILES-derived global features through a dual-level attention module (AMA). The authors claim that this module adaptively fuses the two modalities and achieves state-of-the-art ROC-AUC on Tox21, SIDER, and MUV in 5-shot and 10-shot settings, and they present an MCTS-based rationale extraction method for interpretability.

Significance. If the claims are substantiated, AdaptMol would offer a strong empirical contribution to few-shot drug discovery, and the rationale-extraction approach adds a useful interpretability dimension. The manuscript includes extensive experiments across multiple benchmarks, a cross-domain TDC study, and an ablation analysis, which are valuable. However, the central 'adaptive' mechanism is not implemented as described: the fusion weights in Eqs. (3)-(4) are fixed constants, and the SMILES encoder is not identified, so the reported gains cannot be attributed to input-dependent fusion and are not reproducible. These issues must be resolved before the performance claims can be fully credited.

major comments (4)
  1. [Sec. 3.3, Eqs. (3)-(4)] The so-called adaptive weights beta(g) and beta(s) are constants determined only by the representation level (local vs. global) and by hand-set hyperparameters (k=2, beta_min=0.9, beta_max=1.1 in Appendix D). No term in Eqs. (3)-(9) depends on the input molecule or its extracted features, so the Abstract's claim of 'dynamically integrate' and Section 1's claim of 'assigning higher attention weights to the more informative modality' are not supported by the presented equations. The ablation row labeled 'Adaptive' in Table 4 can only demonstrate the effect of multiplying features by fixed scalars, not of input-dependent fusion. Please either revise the terminology and claims to describe level-dependent weighting, or introduce an input-dependent computation of beta and provide a sensitivity analysis for k, beta_min, and beta_max.
  2. [Sec. 3.3 and Appendix D] The SMILES encoder is described only as 'a large language model' (also in Figure 1(c)); its name, version, pretraining corpus, and fine-tuning status are not reported. This makes the sequence features unreproducible and raises a potential contamination risk: Tox21, SIDER, and MUV are standard benchmarks that may appear in the pretraining corpora of public LLMs, which could inflate the reported gains relative to baselines that do not use such a pretrained encoder. Please specify the exact model and checkpoint, state whether it is frozen or fine-tuned, and discuss or evaluate the overlap between its pretraining data and the evaluation benchmarks.
  3. [Sec. 4.2, Table 1] The text states that AdaptMol 'consistently achieved state-of-the-art performance across different datasets,' but Table 1 shows that on MUV 10-shot, UniMatch achieves 79.40% while AdaptMol achieves 77.16%. The abstract's 'in most cases' is accurate, but the main-text claim is contradicted by the paper's own results and should be corrected. Please also clarify how baseline numbers were obtained (reimplemented or taken from prior papers) and why some entries, such as UniMatch 5-shot, are missing.
  4. [Sec. 3.3, Eq. (6)] The multi-head self-attention layer receives an input Fl_input of dimension N x (d_g + d_a) but is claimed to output Attn_local of dimension N x d_g. Without an explicit projection layer or output dimension specification, this dimension reduction is mathematically undefined. Please provide the exact dimensional transformations for the AMA module, or clarify the projection used in the MultiHead operation.
minor comments (8)
  1. [Table 1 header] The header 'Moldel' should be 'Model'.
  2. [Sec. 4.5, Table 4 title] The ablation study title refers to 'DMA' but the module is called AMA; this typo should be fixed.
  3. [Sec. 4.1] The word 'utlizing' should be 'utilizing'.
  4. [Sec. 2] The word 'moleculer' should be 'molecular'.
  5. [Appendix B.1] The appendix states that four MoleculeNet datasets are used, but Table 1 includes only Tox21, SIDER, and MUV; BACE is used only in the interpretation case study. Please clarify this distinction.
  6. [References] Reference [42] is mislabeled: it points to the TPN paper (same as [38]) instead of the PAR paper (Liu et al., 'Property-aware relation networks for few-shot molecular property prediction,' NeurIPS 2021). The correct citation should be provided.
  7. [Appendix D] The PCA dimension d_a and the MCTS constant cpuct are not reported; please list these hyperparameters for reproducibility.
  8. [Sec. 3.4] The threshold delta_i for accepting rationales is not defined; please state how it is set and whether it varies per property.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: AdaptMol's central claim is an empirical benchmark comparison against external baselines, not a derivation from its own assumptions.

full rationale

The paper's central claim—state-of-the-art ROC-AUC on Tox21, SIDER, and MUV under 5-shot and 10-shot settings—is validated against external baselines in Table 1, so it does not reduce to the model's own definitions or fitted parameters. Equations (1)-(11) specify the architecture and training objective, and none of the reported predictions is constructed from the benchmark labels used to fit a parameter. The 'adaptive' weights in Equations (3)-(4) are fixed constants k=2, beta_min=0.9, beta_max=1.1 (Appendix D), which conflicts with the abstract's 'dynamically integrate' wording; however, this is a specification/implementation discrepancy, not circular reasoning. Appendix F explicitly acknowledges that the fusion mechanism is 'relatively simplistic' and defers data-driven fusion to future work, corroborating that the adaptive-fusion claim is under-supported rather than circular. The interpretability case study (Section 4.4) uses AdaptMol's own prediction scores to select rationales via MCTS; since rationales are defined as subgraphs satisfying ri(Si)>=delta_i, reporting those scores is a consistency check rather than independent evidence, and it is not used to derive the SOTA claim. The unidentified pretrained SMILES encoder (Section 3.3) raises a possible data-contamination concern for external validity, but that is a correctness risk, not circularity. Self-citations such as APN [16] appear as baselines for comparison rather than load-bearing justification. No equation was found that is equivalent to another by construction, and no fitted value is renamed as a prediction.

Assumptions & free parameters 6 free parameters · 3 assumptions · 0 invented entities

The central performance claim rests on standard few-shot machinery and on the untested assumption that fixed beta constants improve fusion. No new physical or chemical entities are introduced.

free parameters (6)
  • k (scaling factor) = 2
    Used in Equations 3-4 to determine beta weights; set in Appendix D without sensitivity analysis.
  • beta_min = 0.9
    Lower bound for fusion weights in Equations 3-4; hand-set in Appendix D.
  • beta_max = 1.1
    Upper bound for fusion weights in Equations 3-4; hand-set in Appendix D.
  • Learning rate range = 0.0005 to 0.05
    Reported as a range in Appendix D; exact value per dataset not given.
  • PCA dimension da = Unspecified
    Dimension of reduced SMILES features in Equation 2 is not stated.
  • MCTS cpuct = Unspecified
    Exploration constant in the MCTS selection formula is not given.
assumptions (3)
  • standard math Multi-head self-attention and prototype networks are effective for few-shot molecular prediction.
    The model builds on these established techniques without re-deriving them (Section 3.3).
  • domain assumption SMILES sequence and molecular graph provide complementary information for molecular property prediction.
    The entire fusion strategy assumes this complementarity (Introduction).
  • ad hoc to paper The fixed beta weight formula (Eq. 3-4) with hand-set constants appropriately balances the two modalities across levels.
    The formula is introduced specifically for AdaptMol and is not derived or shown to be input-dependent (Section 3.3, Appendix D).

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaptMol: Adaptive Fusion from Sequence String to Topological Structure for Few-shot Drug Discovery." pith.science (2026). https://pith.science/paper/Z32ZMMRD

@misc{pith2026250511878,
  author       = {Pith},
  title        = {Pith review of: AdaptMol: Adaptive Fusion from Sequence String to Topological Structure for Few-shot Drug Discovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z32ZMMRD}},
  note         = {Machine review of arXiv:2505.11878}
}
read the original abstract

Accurate molecular property prediction (MPP) is a critical step in modern drug development. However, the scarcity of experimental validation data poses a significant challenge to AI-driven research paradigms. Under few-shot learning scenarios, the quality of molecular representations directly dictates the theoretical upper limit of model performance. We present AdaptMol, a prototypical network integrating Adaptive multimodal fusion for Molecular representation. This framework employs a dual-level attention mechanism to dynamically integrate global and local molecular features derived from two modalities: SMILES sequences and molecular graphs. (1) At the local level, structural features such as atomic interactions and substructures are extracted from molecular graphs, emphasizing fine-grained topological information; (2) At the global level, the SMILES sequence provides a holistic representation of the molecule. To validate the necessity of multimodal adaptive fusion, we propose an interpretable approach based on identifying molecular active substructures to demonstrate that multimodal adaptive fusion can efficiently represent molecules. Extensive experiments on three commonly used benchmarks under 5-shot and 10-shot settings demonstrate that AdaptMol achieves state-of-the-art performance in most cases. The rationale-extracted method guides the fusion of two modalities and highlights the importance of both modalities.

Figures

Figures reproduced from arXiv: 2505.11878 by the authors.

Figure 1
Figure 1. (a) Overview of the proposed AdaptMol framework, where we plot a 2-way 2-shot task. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the Monte Carlo Tree Search (MCTS) method for deriving chemical structure rationales (graph substructures) associ￾ated with high predicted molecular activity. 1. Forward Propagation Select a path from the root s0 to a leaf sL (|sL| ≤ Ns) and evaluate ri(sL). At each state sk, select the bond deletion ak as: ak = arg max a Q(sk, a) + U(sk, a), U(sk, a) = cpuctR(sk, a) s P b N(sk, b) 1 + N(sk, a) . Her… view at source ↗
Figure 3
Figure 3. Using AdaptMol as the scorer for (a) and (b), and single GIN as the scorer for (c) and [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (a) ROC-AUC performance from Tox21 datasets. (b) ROC-AUC performance from SIDER [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: The additional performance of all compared methods on three tasks with a support set of [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 35 canonical work pages

  1. [1]

    Strategies for pre-training graph neural networks,

    W. Hu, B. Liu, J. Gomes, M. Zitnik, P. Liang, V . Pande, and J. Leskovec, “Strategies for pre-training graph neural networks,” in International Conference on Learning Representations (ICLR), 2020

    work page 2020

  2. [2]

    Accurate prediction of molecular properties and drug targets using a self-supervised image representation learning framework,

    X. Zeng, H. Xiang, L. Yu, J. Wang, K. Li, R. Nussinov, and F. Cheng, “Accurate prediction of molecular properties and drug targets using a self-supervised image representation learning framework,” Nature Machine Intelligence, vol. 4, no. 11, pp. 1004–1016, 2022

    work page 2022

  3. [3]

    Dual-view learning based on images and sequences for molecular property prediction,

    X. Zhang, H. Xiang, X. Yang, J. Dong, X. Fu, X. Zeng, H. Chen, and K. Li, “Dual-view learning based on images and sequences for molecular property prediction,” IEEE Journal of Biomedical and Health Informatics, 2023

    work page 2023

  4. [4]

    Similarity maps—a visualization strategy for molecular fingerprints and machine-learning methods,

    S. Riniker and G. A. Landrum, “Similarity maps—a visualization strategy for molecular fingerprints and machine-learning methods,” Journal of Cheminformatics, pp. 1–7, 2013

    work page 2013

  5. [5]

    Computational methods in drug discovery,

    G. Sliwoski, S. Kothiwale, J. Meiler, and E. W. Lowe, “Computational methods in drug discovery,” Pharmacological Reviews, pp. 334–395, 2014

    work page 2014

  6. [6]

    Communicative representation learning on attributed molecular graphs,

    Y . Song, S. Zheng, Z.-m. Niu, Z.-H. Fu, Y . Lu, and Y . Yang, “Communicative representation learning on attributed molecular graphs,” in Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI), 2020, pp. 2831–2838

    work page 2020

  7. [7]

    Molecular contrastive learning with chemical element knowledge graph,

    Y . Fang, Q. Zhang, H. Yang, X. Zhuang, S. Deng, W. Zhang, M. Qin, Z. Chen, X. Fan, and H. Chen, “Molecular contrastive learning with chemical element knowledge graph,” inProceedings of the 36th AAAI Conference, 2022, pp. 3968–3976

    work page 2022

  8. [8]

    Knowledge graph- enhanced molecular contrastive learning with functional prompt,

    Y . Fang, Q. Zhang, N. Zhang, Z. Chen, X. Zhuang, X. Shao, X. Fan, and H. Chen, “Knowledge graph- enhanced molecular contrastive learning with functional prompt,” Nature Machine Intelligence, pp. 1–12, 2023. 9

    work page 2023

Show all 47 references
  1. [9]

    Learning feed-forward one-shot learners,

    L. Bertinetto, J. F. Henriques, J. Valmadre, P. H. Torr, and A. Vedaldi, “Learning feed-forward one-shot learners,” in Advances in Neural Information Processing Systems (NIPS), 2016, pp. 523–531

    2016

  2. [10]

    Low data drug discovery with one-shot learning,

    H. Altae-Tran, B. Ramsundar, A. S. Pappu, and V . Pande, “Low data drug discovery with one-shot learning,” ACS Central Science, vol. 3, no. 4, pp. 283–293, 2017

    2017

  3. [11]

    Meta learning with graph attention networks for low-data drug discovery,

    Q. Lv, G. Chen, Z. Yanget al., “Meta learning with graph attention networks for low-data drug discovery,” IEEE Transactions on Neural Networks and Learning Systems, vol. 35, pp. 11 218–11 230, 2024

    2024

  4. [12]

    Attribute-guided prototype network for few-shot molecular property prediction,

    L. Hou, H. Xiang, X. Zeng, D. Cao, L. Zeng, and B. Song, “Attribute-guided prototype network for few-shot molecular property prediction,” Briefings in Bioinformatics, vol. 25, no. 5, p. bbae394, 08 2024. [Online]. Available: https://doi.org/10.1093/bib/bbae394

    2024 doi

  5. [13]

    Fp-gnn: a versatile deep learning architecture for enhanced molecular property prediction,

    H. Cai, H. Zhang, D. Zhao, J. Wu, and L. Wang, “Fp-gnn: a versatile deep learning architecture for enhanced molecular property prediction,” Briefings in Bioinformatics, vol. 23, no. 6, p. bbac408, 2022. [Online]. Available: https://academic.oup.com/bib/advance-article/doi/10.1...

    2022 doi

  6. [14]

    Smiclr: Contrastive learning on multiple molecu- lar representations for semisupervised and unsupervised representation learning,

    G. A. Pinheiro, J. L. F. Da Silva, and M. G. Quiles, “Smiclr: Contrastive learning on multiple molecu- lar representations for semisupervised and unsupervised representation learning,” Journal of Chemical Information and Modeling, vol. 62, no. 17, pp. 3948–3960, 2022

    2022

  7. [15]

    Improving molecular pretraining with complementary featurizations,

    Y . Zhu, D. Chen, Y . Duet al., “Improving molecular pretraining with complementary featurizations,”arXiv preprint arXiv:2209.15101, 2022

    2022 arXiv

  8. [16]

    Attribute-guided prototype network for few-shot molecular property prediction,

    L. Hou, H. Xiang, X. Zeng, D. Cao, L. Zeng, and B. Song, “Attribute-guided prototype network for few-shot molecular property prediction,” Briefings in Bioinformatics, vol. 25, no. 5, p. bbae394, Jul 2024

    2024

  9. [17]

    Matching networks for one shot learning,

    O. Vinyals, C. Blundell, T. Lillicrap, K. Kavukcuoglu, and D. Wierstra, “Matching networks for one shot learning,” in Advances in Neural Information Processing Systems (NIPS), 2016, pp. 3630–3638

    2016

  10. [18]

    A closer look at few-shot classification,

    W.-Y . Chen, Y .-C. Liu, Z. Kira, Y .-C. F. Wang, and J.-B. Huang, “A closer look at few-shot classification,” in International Conference on Learning Representations (ICLR), 2019

    2019

  11. [19]

    A systematic review on data scarcity problem in deep learning: Solution and applications,

    A. Bansal, R. Sharma, and M. Kathuria, “A systematic review on data scarcity problem in deep learning: Solution and applications,” ACM Computing Surveys, vol. 54, no. 10s, pp. 1–29, 2022. [Online]. Available: https://doi.org/10.1145/3502287

    2022 doi

  12. [20]

    Adaptive task sampling for meta-learning,

    C. Liu, Z. Wang, D. Sahoo, Y . Fang, K. Zhang, and S. C. H. Hoi, “Adaptive task sampling for meta-learning,” in Computer Vision – ECCV 2020. Springer, 2020, pp. 752–769. [Online]. Available: https://doi.org/10.1007/978-3-030-58523-5_44

    2020 doi

  13. [21]

    Meta-learning with an adaptive task scheduler,

    H. Yao, Y . Wang, Y . Wei, P. Zhao, M. Mahdavi, D. Lian, and C. Finn, “Meta-learning with an adaptive task scheduler,” in Advances in Neural Information Processing Systems, vol. 34, 2021, pp. 7497–7509. [Online]. Available: https://arxiv.org/abs/2110.14057

    2021 arXiv

  14. [22]

    Meta-learning in neural networks: A survey,

    T. M. Hospedales, A. Antoniou, P. Micaelli, and A. J. Storkey, “Meta-learning in neural networks: A survey,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 9, pp. 5149–5169,

  15. [23]

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

    C. Finn, P. Abbeel, and S. Levine, “Model-agnostic meta-learning for fast adaptation of deep networks,” in Proceedings of the 34th International Conference on Machine Learning , ser. Proceedings of Machine Learning Research, D. Precup and Y . W. Teh, Eds., vol. 70. PMLR, 06–11...

    2017

  16. [24]

    Prototypical networks for few-shot learning,

    J. Snell, K. Swersky, and R. S. Zemel, “Prototypical networks for few-shot learning,” in Advances in Neural Information Processing Systems , vol. 30, 2017, pp. 4077–4087. [Online]. Available: https://papers.nips.cc/paper/6996-prototypical-networks-for-few-shot-learning

    2017

  17. [25]

    Property-aware relation networks for few-shot molecular property prediction,

    Y . Wang, A. Abuduweili, Q. Yaoet al., “Property-aware relation networks for few-shot molecular property prediction,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 34, 2021, pp. 17 441– 17 454

    2021

  18. [26]

    Few-shot graph learning for molecular property prediction,

    Z. Guo, C. Zhang, W. Yu, J. Herr, O. Wiest, M. Jiang, and N. V . Chawla, “Few-shot graph learning for molecular property prediction,” in Proceedings of the 30th International Conference on World Wide Web (WWW), 2021, pp. 2559–2567

    2021

  19. [27]

    Low data drug discovery with one-shot learning,

    H. Altae-Tran, B. Ramsundar, A. S. Pappu et al., “Low data drug discovery with one-shot learning,”ACS Central Science, vol. 3, no. 4, pp. 283–293, 2017

    2017

  20. [28]

    Few-shot learning for low-data drug discovery,

    D. Vella and J.-P. Ebejer, “Few-shot learning for low-data drug discovery,” Journal of Chemical Information and Modeling , vol. 63, no. 1, pp. 27–42, 2022. [Online]. Available: https://doi.org/10.1021/acs.jcim.2c00779

    2022 doi

  21. [29]

    Stop explaining black box machine learning models for high stakes decisions and use inter- pretable models instead,

    C. Rudin, “Stop explaining black box machine learning models for high stakes decisions and use inter- pretable models instead,” Nature Machine Intelligence, vol. 1, no. 5, pp. 206–215, 2019. 10

    2019

  22. [30]

    “why should i trust you?

    M. T. Ribeiro, S. Singh, and C. Guestrin, ““why should i trust you?”: Explaining the predictions of any classifier,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York, NY , USA: Association for Computing Machinery, ...

    2016

  23. [31]

    Axiomatic attribution for deep networks,

    M. Sundararajan, A. Taly, and Q. Yan, “Axiomatic attribution for deep networks,” inProceedings of the 34th International Conference on Machine Learning, vol. 70. PMLR, 2017, pp. 3319–3328. [Online]. Available: https://proceedings.mlr.press/v70/sundararajan17a.html

    2017

  24. [32]

    Multi-objective molecule generation using interpretable substructures,

    W. Jin, R. Barzilay, and T. Jaakkola, “Multi-objective molecule generation using interpretable substructures,” in Proceedings of the 37th International Conference on Machine Learning (ICML) . PMLR, 2020, pp. 4849–4859. [Online]. Available: https://proceedings.mlr.press/v119/jin20b

    2020

  25. [33]

    Mastering the game of go without human knowledge,

    D. Silver, J. Schrittwieser, K. Simonyan, I. Antonoglou, A. Huang, A. Guez, T. Hubert, L. Baker, M. Lai, A. Bolton et al., “Mastering the game of go without human knowledge,”Nature, vol. 550, no. 7676, pp. 354–359, 2017

    2017

  26. [34]

    Moleculenet: A benchmark for molecular machine learning,

    Z. Wu, B. Ramsundar, E. N. Feinberg, J. Gomes, C. Geniesse, A. S. Pappu, K. Leswing, and V . S. Pande, “Moleculenet: A benchmark for molecular machine learning,”Chemical Science, vol. 9, no. 2, pp. 513–530,

  27. [35]

    Siamese neural networks for one-shot image recognition,

    G. Koch, R. Zemel, and R. Salakhutdinov, “Siamese neural networks for one-shot image recognition,” in ICML Deep Learning Workshop , vol. 2. Lille, France: PMLR, 2015. [Online]. Available: https://www.cs.cmu.edu/~rsalakhu/papers/oneshot1.pdf

    2015

  28. [36]

    Attention-based bidirectional long short-term memory networks for relation classification,

    Z. Lin, M. Feng, C. N. d. Santos, M. Yu, B. Xiang, B. Zhou, and Y . Bengio, “Attention-based bidirectional long short-term memory networks for relation classification,” in Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (ACL) . Associati...

    2016

  29. [37]

    Cross-domain few-shot learning by representation fusion,

    T. Adler, J. Brandstetter, M. Widrich, A. Mayr, D. P. Kreil, M. Kopp, G. Klambauer, and S. Hochreiter, “Cross-domain few-shot learning by representation fusion,” CoRR, vol. abs/2010.06498, 2020. [Online]. Available: https://arxiv.org/abs/2010.06498

    2010 arXiv

  30. [38]

    Learning to propagate labels: Transductive propagation network for few-shot learning,

    Y . Liu, J. Lee, M. Park, S. Kim, E. Yang, S. J. Hwang, and Y . Yang, “Learning to propagate labels: Transductive propagation network for few-shot learning,”arXiv preprint arXiv:1805.10002, 2018

    2018 arXiv

  31. [39]

    Boil: Towards representation change for few-shot learning,

    J. Oh, H. Yoo, C. Kim, and S.-Y . Yun, “Boil: Towards representation change for few-shot learning,” 2021. [Online]. Available: https://arxiv.org/abs/2008.08882

    2021 arXiv

  32. [40]

    Edge-labeling graph neural network for few-shot learning,

    J. Kim, T. Kim, S. Kim, and C. D. Yoo, “Edge-labeling graph neural network for few-shot learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 11–20

    2019

  33. [41]

    Unimatch: Universal matching from atom to task for few-shot drug discovery,

    R. Li, M. Li, W. Liu, Y . Zhou, X. Zhou, Y . Yao, Q. Zhang, and H. Chen, “Unimatch: Universal matching from atom to task for few-shot drug discovery,” 2025. [Online]. Available: https://arxiv.org/abs/2502.12453

    2025 arXiv

  34. [42]

    Learning to propagate labels: Transduc- tive propagation network for few-shot learning,

    Y . Liu, J. Lee, M. Park, S. Kim, E. Yang, S. J. Hwang, and Y . Yang, “Learning to propagate labels: Transduc- tive propagation network for few-shot learning,” in International Conference on Learning Representations (ICLR), 2019

    2019

  35. [43]

    Therapeutics data commons: Machine learning datasets and tasks for drug discovery and development,

    K. Huang, T. Fu, W. Gao, Y . Zhao, Y . Roohani, J. Leskovec, C. W. Coley, C. Xiao, J. Sun, and M. Zitnik, “Therapeutics data commons: Machine learning datasets and tasks for drug discovery and development,”

  36. [44]

    Adam: a method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: a method for stochastic optimization,”arXiv preprint arXiv:1412.6980, 2014. 11 Appendices A Algorithm In order to clearly describe the training process of AdaptMol framework, we show the process in Algorithm 1. Algorithm 1 Meta-training procedure...

    2014 arXiv

  37. [2018]

    Available: https://pubs.rsc.org/en/content/articlelanding/2018/sc/c7sc02664a

    [Online]. Available: https://pubs.rsc.org/en/content/articlelanding/2018/sc/c7sc02664a

    2018

  38. [2021]

    Available: https://arxiv.org/abs/2102.09548

    [Online]. Available: https://arxiv.org/abs/2102.09548

    arXiv

  39. [2022]

    Available: https://doi.org/10.1109/TPAMI.2021.3079209

    [Online]. Available: https://doi.org/10.1109/TPAMI.2021.3079209

    2021

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.