Pith. sign in

REVIEW 4 major objections 4 minor 52 references

Beyond Gene Reconstruction: Learning Cell Representations through Complementary Transcriptomic Views

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Cell-level contrastive pretraining on complementary gene views produces more discriminative, transferable single-cell representations than masked reconstruction alone.

desk verdict A solid, honest empirical recipe for contrastive single-cell pretraining; the transfer claim needs a data-overlap audit and multi-seed runs before it is fully convincing. read the letter →

arxiv 2608.00985 v1 pith:54KQMRUO submitted 2026-08-02 cs.LG cs.AIq-bio.GN

Jiaqi Xiong , Yuntao hu , Yu Zheng , Yifei Shi , Xinyue Guo , Jiaxin Qi This is my paper · ORCID
classification cs.LGcs.AIq-bio.GN
keywords single-celltranscriptomicscontrastivelearningcellrepresentationmaskedexpressionreconstructionfoundationmodelscell-typeannotationgeneregulatorynetworkinferencepretrainingobjective
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

Single-cell foundation models are pretrained mainly by reconstructing masked expression values, an objective that supervises individual genes but says nothing about the geometry of the whole-cell embedding. This paper argues that a cell-level contrastive objective, applied the right way, fills that gap. The method, CoCoS, gives each cell two complementary views by splitting its genes along co-expression structure, builds hard negatives that permute expression values while keeping gene identities fixed, and switches the contrastive loss on only after a label-free readiness check on held-out cells. With these pieces, the CoCoS-GPC embeddings beat reconstruction-only references on cell-type annotation (74.82% vs 70.99% mean k-NN against scGPT, 78.85% vs 77.02% linear probe) and give the highest mean AUROC/AUPRC point estimates on six gene-regulatory networks. If the paper is right, a bounded contrastive term on complementary views is a viable complement to masked reconstruction in single-cell pretraining.

What carries the argument

The load-bearing mechanism is the complementary view pair. A weighted gene co-occurrence graph is bipartitioned so strongly co-occurring genes land in opposite panels; each cell is masked twice, once per panel, so every gene value is observed in one view and masked in the other. This yields a biologically anchored positive pair while keeping the reconstruction objective intact. Around that pair, CoCoS builds a contrast set of same-view and cross-view in-batch cells plus $K=4$ fixed-identity value derangements, values permuted but gene identities and masking preserved, forcing the encoder to use gene–value correspondence rather than gene-set composition. A competence gate keeps the contrastiv

What would settle it

An overlap audit: compare the 1,799,684 training rows against the cell rows used in the ten cell-type annotation datasets and the six regulatory-network evaluation sets. If a substantial fraction of downstream cells or their exact transcriptomic profiles appear verbatim in the training pool, the reported k-NN and GRN gains would be partly memorization rather than transfer, weakening the central claim; if overlap is negligible, the transfer conclusion stands. A secondary check is to run the K=0 versus K=4 shortcut diagnostic for the full 140,600-update training rather than 1,000 updates, testin

Watch

Extended reading notes

Core claim

The paper claims that masked expression-value reconstruction alone leaves whole-cell representation geometry unsupervised, and that a properly bounded contrastive objective over complementary transcriptomic views closes that gap. CoCoS keeps every measured gene–value pair intact and instead partitions each cell's genes by co-expression into two complementary views; it builds contrast sets whose hardest negatives are fixed-identity value derangements, so gene-set composition cannot separate positives from negatives; and it leaves the contrastive weight at zero until label-free probes on held-out cells show stable cross-view concordance, low reconstruction error, and preserved embedding divers

Load-bearing premise

The transfer claim presupposes that the ten downstream cell-type datasets and six gene-regulatory networks are not materially present in the 1.8-million-row pretraining pool built from the same public single-cell atlas; the paper guarantees only that downstream labels never enter pretraining, not that the same cells never appear in both.

Editorial extensions

If this is right

  • If CoCoS-GPC is right, single-cell foundation-model pretraining should include a bounded cell-level contrastive objective alongside masked reconstruction: frozen whole-cell embeddings transfer better to cell-type annotation than the reconstruction-only reference and scGPT (74.82% vs 70.99% mean k-NN).
  • The fixed-identity derangement diagnostic shows a measurable shortcut: without them, confidence against identity-matched negatives sits at 0.639 with an ordinary–identity gap of 0.358; with $K=4$ these move to 0.969 and 0.028, so the contrast objective learns gene–value correspondence, not gene-set composition.
  • Gating contrastive onset on label-free competence beats fixed schedules: the GPC trigger at step 93,000 improves over immediate contrast by 1.50 k-NN points and over both one- and two-epoch warmups, making data-dependent onset a load-bearing design choice.
  • The same frozen encoder supports gene-regulatory edge prediction, with the highest mean AUROC (0.8733) and AUPRC (0.7120) point estimates among compared variants on six networks, although per-network winners differ.
  • Because both views share one encoder and one optimizer update per minibatch, the contrastive addition is compatible with a modest training budget (10 epochs, one RTX 4090), making the recipe practical to reproduce.

Reading between the lines

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

  • My inference: the label-free competence gate is a general pattern. Any pretraining run that couples a stabilizing objective with a potentially corrupting one could trigger the second objective using the same held-out readiness probes, not just single-cell contrast.
  • My inference: because each positive pair is two views of the same measured cell rather than a perturbed copy, the design sidesteps the open question of which expression perturbations preserve cell identity; the same view-construction idea could transfer to other contrastive methods on sparse, high-dimensional molecular data.
  • My inference: the paper reports one training seed and point estimates, so the reported margins are not yet distributional claims; a multi-seed rerun of the ten-dataset protocol would show whether the k-NN gain is stable or within noise, especially where per-network GRN winners vary.
  • My inference: a natural scale-up test follows from the paper's own numbers. Expression-aware sampling shifted the onset from 96,000 to 93,000 steps and improved accuracy by 0.88 k-NN points, suggesting the sampler and the gate interact; ablating them on larger corpora would show whether that interaction grows or saturates.
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 / 4 minor

Summary. The paper proposes CoCoS, a contrastive pretraining framework for single-cell transcriptomics that augments masked expression reconstruction with three components: co-expression-guided gene partitioning into two complementary views, expression-aware contrast-set construction with fixed-identity value derangements, and competence-gated contrastive onset. The authors evaluate frozen embeddings on ten cell-type annotation datasets and six BEELINE GRN networks, reporting mean k-NN and linear-probe accuracies as well as AUROC/AUPRC point estimates. The central claim is that properly bounded whole-cell contrastive supervision yields more discriminative and transferable cell representations than reconstruction alone. The paper is candid about limitations, including one training seed, point-estimate GRN results, and a short-fork diagnostic that does not measure downstream performance.

Significance. If the central claim holds, CoCoS offers a practical recipe for adding cell-level contrastive supervision to masked single-cell pretraining, addressing two real transcriptome-specific challenges: constructing biologically valid positive pairs and avoiding gene-identity shortcuts. The method is principled, and the paper includes useful controlled diagnostics: the fixed-identity shortcut experiment in Figure 3 is well designed, and the same-architecture scBenchmark baseline provides a more trustworthy comparison than the external scGPT checkpoint. The manuscript is also unusually transparent about its limitations. However, the transfer claim currently rests on an unverified data-disjointness assumption and on single-seed point estimates, and the balanced-random panel control nearly matches the full method, weakening the specific case for co-expression-guided partitioning. These issues are fixable and do not invalidate the framework, but they must be addressed before the broad transferability conclusion is established.

major comments (4)
  1. [Datasets / Downstream evaluation] The transfer conclusion presupposes that the downstream evaluation cells are not in the 1,813,780-row pretraining corpus. The paper says only that annotation and GRN labels do not enter pretraining; it does not audit cell overlap. Both the pretraining corpus and the ten Qi et al. cell-type datasets are drawn from the CELLxGENE ecosystem, and the BEELINE GRN atlases plausibly overlap it. Exact or near overlap would let contrastive pretraining memorize cell identities and inflate the reported gains; the margins (+3.83 k-NN over scGPT, +0.88 over uniform sampling) are small enough for such artifacts to matter. Please add an overlap audit (e.g., barcode or expression-profile deduplication) for all downstream sets, or restrict the claim to non-transfer performance.
  2. [Experimental Setup; Tables 1-3] All main results use one pretraining seed (42) and one fixed downstream split. The paper's central comparative claim is therefore supported only by point estimates without uncertainty or significance. For example, Table 3 differences between balanced-random and proposed (0.48 k-NN, 0.25 linear) and between uniform and expression-aware sampling (0.88 k-NN, 0.25 linear) are within plausible training noise, and the +3.83 k-NN margin over scGPT could be seed-specific. The split-seed 42-44 reevaluation varies only the downstream split for a fixed checkpoint and does not address pretraining variance. Please provide multiple pretraining seeds (or paired bootstrap) for the main tables.
  3. [Method, Eq. (2)] The partition objective in Eq. (2) maximizes the total co-occurrence weight of cross-view gene pairs, but the algorithm that follows assigns each gene opposite the panel containing more of its already-assigned neighbors, counting each neighbor once 'irrespective of w_gh.' Edge weights therefore enter only through the visit order, not through the placement decision, so the method does not actually optimize Eq. (2). If the count-based heuristic is intentional, it should be stated as an approximation, and the claim that views are 'co-expression-guided' should be tied to the heuristic actually used. As written, the formal objective and the implemented procedure are inconsistent.
  4. [Table 3] The balanced-random panel control is nearly as strong as the full method (74.34 vs. 74.82 k-NN; 78.44 vs. 78.85 linear). Since this control differs only in how the two panels are chosen, Table 3 does not demonstrate that co-expression-guided partitioning contributes materially to the proposed system. Given that the first contribution is 'co-expression-guided gene partitioning,' the paper should report an ablation that isolates the partition choice while holding all other components (including the flip probability and GPC triggering) fixed, and state the size of the partition effect.
minor comments (4)
  1. [Experimental Setup] The description of Geneformer contains an incomplete cross-reference: 'as detailed in Section.' with no section number.
  2. [Method, Eq. (4)] The repertoire score 'q_p = ||std(Z_p)||_2 / p_erank_p' is ambiguous: define p and 'erank' explicitly, and clarify how Z_p is normalized before computing the standard deviation.
  3. [Table 4] The GPC row is described as an unmatched reference, but the surrounding text tends to read as a comparison with the fixed-onset controls. Please state explicitly that the GPC configuration differs in multiple factors and therefore does not by itself establish superiority of the data-dependent onset.
  4. [Reevaluation paragraph] The text reports 74.87 ± 0.06 mean k-NN over downstream split seeds 42-44, while Table 1 reports 74.82 for seed 42. Clarify whether the checkpoint, evaluator, or averaging procedure differs, or correct the discrepancy.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method's gains are not forced by construction or by self-citation.

full rationale

This is an empirical method paper. The contrastive objective (Eq. 3) is a standard InfoNCE loss applied to complementary views; the views are constructed from a co-expression partition (Eq. 2), not from the downstream labels. The competence-gated onset uses a held-out, label-free sentinel cohort and is calibrated on early probes; it is a training-schedule decision, not a fitted parameter that predicts the reported metrics. The paper explicitly treats the GPC row as an unmatched contextual reference (Table 4), and the fixed-identity shortcut diagnostic is a controlled fork that does not estimate downstream performance. Self-citations (Xiong et al. 2025; Qi et al. 2025) appear in motivation and in the description of the benchmark/backbone, but they are not used to justify the central claim; the evaluation is against fixed external baselines and public datasets. The main risk—possible overlap between the pretraining corpus and downstream evaluation cells—is a data-leakage concern, not a circular-derivation concern, and the paper does not claim an overlap audit. No equation reduces to another by definition, and no fitted parameter is renamed as a prediction. Therefore the derivation chain is self-contained and no circularity is present.

Assumptions & free parameters 9 free parameters · 5 assumptions · 0 invented entities

The central claim depends on hand-selected or sentinel-calibrated hyperparameters (contrastive weight, ramp, K, mask ratio, readiness thresholds) and on domain assumptions that co-expression partitions and value permutations create valid positive and negative views. No new physical entities are introduced.

free parameters (9)
  • beta_max = not reported
    Maximum weight of the contrastive loss after the competence gate triggers; chosen by the authors and not specified in the main text.
  • contrastive ramp duration R = not reported
    Determines how quickly beta(t) rises from 0 to beta_max after the trigger; value not given in main text.
  • K (fixed-identity hard negatives per anchor) = 4
    Number of value-derangement negatives used in the proposed configuration; selected by design and varied in the diagnostic.
  • mask ratio = 40%
    Fraction of expression values masked in each view; chosen by hand.
  • observed-gene filter and per-view cap = min 50, cap 512
    Cells with fewer than 50 observed genes are excluded and each view is capped at 512 genes; chosen for computational and sparsity reasons.
  • panel-assignment flip probability = 5%
    Probability that a gene is reassigned to the opposite view in a minibatch; chosen by hand.
  • co-occurrence graph neighbor truncation = not reported
    Each gene retains its strongest neighbors before the bipartition, but the number or threshold of retained neighbors is not specified.
  • readiness thresholds (concordance, slope, reconstruction reference, repertoire reference) = calibrated on early sentinel probes
    Calibration values determine when the contrastive objective is activated; reported only in supplementary material.
  • consecutive probes required = 3
    Number of consecutive stable probes before readiness is declared; chosen by hand.
assumptions (5)
  • domain assumption A Transformer encoder with gene-identity tokens and continuous expression values can learn useful gene dependencies from masked expression reconstruction.
    Backbone is imported from scBenchmark and used throughout; the paper does not question this architectural choice.
  • domain assumption Co-occurrence of genes among highly expressed genes is a valid proxy for shared biological programs, so placing co-occurring genes in opposite panels yields complementary but biologically aligned views.
    Equation 2 and the partitioning method in the Method section; this is the core premise of the complementary-view design.
  • domain assumption Permuting expression values while keeping gene identities fixed creates valid hard negatives that share gene-set composition but differ in biological state.
    Expression-Aware Contrast-Set Construction; if the permutation preserved cell identity, the contrast objective would be mis-specified.
  • ad hoc to paper The label-free readiness signals (GPC, reconstruction error, repertoire) are a valid proxy for when contrastive learning will improve downstream transfer.
    Competence-Gated Contrastive Onset; the criteria are chosen by the authors and calibrated on a sentinel cohort, with no independent validation that they predict downstream accuracy.
  • standard math The InfoNCE contrastive loss with in-batch negatives promotes discriminative whole-cell embeddings in the single-cell setting.
    Equation 3; standard contrastive-learning assumption, adapted here with same-view and cross-view negatives.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Gene Reconstruction: Learning Cell Representations through Complementary Transcriptomic Views." pith.science (2026). https://pith.science/paper/54KQMRUO

@misc{pith2026260800985,
  author       = {Pith},
  title        = {Pith review of: Beyond Gene Reconstruction: Learning Cell Representations through Complementary Transcriptomic Views},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/54KQMRUO}},
  note         = {Machine review of arXiv:2608.00985}
}
read the original abstract

The rapid growth of single-cell transcriptomic data has enabled the development of foundation models pretrained primarily by reconstructing masked expression values. This objective encourages these models to learn gene dependencies but does not directly optimize whole-cell representations, which are essential for many downstream tasks. To bridge this gap, we propose a contrastive pretraining framework that learns cell representations through complementary transcriptomic views. Since standard contrastive learning is not readily applicable to single-cell pretraining, we introduce specific adaptations along three dimensions --- co-expression-guided gene partitioning, expression-aware contrast-set construction, and competence-gated contrastive onset. Specifically, we first construct two complementary views of each cell by partitioning its genes according to their co-expression structure. Then, to prevent the model from using gene-set identity as a shortcut, we construct hard negatives by permuting expression values while keeping gene identities unchanged. Finally, we introduce a competence-aware controller to determine how the contrastive objective is applied. Experiments on cell-type annotation and gene regulatory network inference demonstrate competitive transfer under the evaluated protocols. In the six-network GRN evaluation, our method records the highest mean AUROC and AUPRC point estimates among the compared variants, while the highest-scoring variant differs across individual networks. These results establish complementary-view contrastive learning as an effective direction for single-cell pretraining beyond gene reconstruction.

Figures

Figures reproduced from arXiv: 2608.00985 by the authors.

Figure 1
Figure 1. Motivation for CoCoS. (a) Masked expression￾value prediction directly supervises reconstruction through gene-level features but does not impose a metric objective on the whole-cell representation. (b) Naively transferring contrastive learning raises two transcriptome-specific chal￾lenges: unconstrained expression perturbations may alter the cellular state used to define a positive pair, while negatives with differen… view at source ↗
Figure 2
Figure 2. Overview of CoCoS-GPC. Co-expression-guided partitioning and expression-aware sampling form two disjoint views, which are masked and reconstructed by a shared Transformer. The paired same-cell views define the positive, while same-view and cross-view in-batch examples together with fixed-identity value derangements form the contrast set. A label-free sentinel tracks smoothed gene-program concordance, reconstruction … view at source ↗
Figure 3
Figure 3. Fixed-identity shortcut diagnostic. From a shared [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 33 canonical work pages

  1. [1]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    A Simple and Comprehensive Benchmark for Single-Cell Transcriptomics , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=. 2025 , doi=

    work page 2025

  2. [2]

    BMC Bioinformatics , volume=

    Cross-Attention Graph Neural Networks for Inferring Gene Regulatory Networks with Skewed Degree Distribution , author=. BMC Bioinformatics , volume=. 2025 , publisher=. doi:10.1186/s12859-025-06186-1 , url=

    work page doi:10.1186/s12859-025-06186-1 2025

  3. [3]

    Nature Methods , volume=

    scGPT: toward building a foundation model for single-cell multi-omics using generative AI , author=. Nature Methods , volume=. 2024 , publisher=

    work page 2024

  4. [4]

    Nature , volume=

    Transfer learning enables predictions in network biology , author=. Nature , volume=. 2023 , publisher=

    2023

  5. [5]

    Nature Machine Intelligence , volume=

    scBERT as a large-scale pretrained deep language model for cell type annotation of single-cell RNA-seq data , author=. Nature Machine Intelligence , volume=. 2022 , publisher=

    2022

  6. [6]

    Nature Methods , volume=

    Large-scale foundation model on single-cell transcriptomics , author=. Nature Methods , volume=. 2024 , publisher=

    2024

  7. [7]

    Hinton , title=

    Ting Chen and Simon Kornblith and Mohammad Norouzi and Geoffrey E. Hinton , title=. Proceedings of the 37th International Conference on Machine Learning,. 2020 , url=

    work page 2020

  8. [8]

    Girshick , title=

    Kaiming He and Haoqi Fan and Yuxin Wu and Saining Xie and Ross B. Girshick , title=. 2020. 2020 , doi=

    work page 2020

Show all 52 references
  1. [9]

    Proceedings of the 38th International Conference on Machine Learning,

    Alec Radford and Jong Wook Kim and Chris Hallacy and Aditya Ramesh and Gabriel Goh and Sandhini Agarwal and Girish Sastry and Amanda Askell and Pamela Mishkin and Jack Clark and Gretchen Krueger and Ilya Sutskever , title=. Proceedings of the 38th International Conference on M...

    2021

  2. [10]

    2018 , eprint=

    Representation Learning with Contrastive Predictive Coding , author=. 2018 , eprint=

    2018

  3. [11]

    Alex Kendall and Yarin Gal and Roberto Cipolla , title=. 2018. 2018 , doi=

    2018

  4. [12]

    GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks , booktitle=

    Zhao Chen and Vijay Badrinarayanan and Chen. GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks , booktitle=. 2018 , url=

    2018

  5. [13]

    Curriculum Learning , booktitle=

    Yoshua Bengio and J. Curriculum Learning , booktitle=. 2009 , doi=

    2009

  6. [14]

    Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages=

    Emmanouil Antonios Platanios and Otilia Stretcu and Graham Neubig and Barnabas Poczos and Tom Mitchell , title=. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies , pages=. 2019 , doi=

    2019

  7. [15]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Yajing Kong and Liu Liu and Jun Wang and Dacheng Tao , title=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=. 2021 , url=

    2021

  8. [16]

    Davison and Edward Johns , title=

    Shikun Liu and Stephen James and Andrew J. Davison and Edward Johns , title=. Transactions on Machine Learning Research , year=

  9. [17]

    Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , year=

    Tianhe Yu and Saurabh Kumar and Abhishek Gupta and Sergey Levine and Karol Hausman and Chelsea Finn , title=. Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , year=

    2020

  10. [18]

    Davison , title=

    Shikun Liu and Edward Johns and Andrew J. Davison , title=. 2019 , doi=

    2019

  11. [19]

    Advances in Neural Information Processing Systems 36 (NeurIPS 2023) , year=

    Bo Liu and Yihao Feng and Peter Stone and Qiang Liu , title=. Advances in Neural Information Processing Systems 36 (NeurIPS 2023) , year=

    2023

  12. [20]

    Le and Tengyu Ma and Adams Wei Yu , title=

    Sang Michael Xie and Hieu Pham and Xuanyi Dong and Nan Du and Hanxiao Liu and Yifeng Lu and Percy Liang and Quoc V. Le and Tengyu Ma and Adams Wei Yu , title=. Advances in Neural Information Processing Systems 36 (NeurIPS 2023) , year=

    2023

  13. [21]

    2024 , eprint=

    LangCell: Language-Cell Pre-training for Cell Identity Understanding , author=. 2024 , eprint=

    2024

  14. [22]

    bioRxiv , year=

    scConcept: Contrastive pretraining for technology-agnostic single-cell representations beyond reconstruction , author=. bioRxiv , year=. doi:10.1101/2025.10.14.682419 , url=

    2025 doi

  15. [23]

    2021 , eprint=

    Contrastive Learning with Hard Negative Samples , author=. 2021 , eprint=

    2021

  16. [24]

    Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , pages=

    Kalantidis, Yannis and Sariyildiz, Mert Bulent and Pion, Noe and Weinzaepfel, Philippe and Larlus, Diane , title=. Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , pages=. 2020 , url=

    2020

  17. [25]

    Advances in Neural Information Processing Systems 34 (NeurIPS 2021) , pages=

    Liu, Bo and Liu, Xingchao and Jin, Xiaojie and Stone, Peter and Liu, Qiang , title=. Advances in Neural Information Processing Systems 34 (NeurIPS 2021) , pages=. 2021 , url=

    2021

  18. [26]

    Proceedings of the 39th International Conference on Machine Learning , pages=

    Multi-Task Learning as a Bargaining Game , author=. Proceedings of the 39th International Conference on Machine Learning , pages=. 2022 , editor=

    2022

  19. [27]

    2023 , eprint=

    Efficient Online Data Mixing For Language Model Pre-Training , author=. 2023 , eprint=

    2023

  20. [28]

    and Liu, Han , year=

    ElSheikh, Ali and Wang, Rui-Xi and Wu, Weimin and Wen, Yibo and Dibaeinia, Payam and Zhang, Jennifer Yuntong and Hu, Jerry Yao-Chieh and Knudson, Mei and Babu, Sudarshan and Sun, Shao-Hua and Khan, Aly A. and Liu, Han , year=. 2602.02093 , archivePrefix=

  21. [29]

    BMC Bioinformatics , volume=

    Contrastive self-supervised clustering of scRNA-seq data , author=. BMC Bioinformatics , volume=. 2021 , doi=

    2021

  22. [30]

    Self-supervised contrastive learning for integrative single cell

    Han, Wenkai and Cheng, Yuqi and Chen, Jiayang and Zhong, Huawen and Hu, Zhihang and Chen, Siyuan and Zong, Licheng and Hong, Liang and Chan, Ting-Fung and King, Irwin and Gao, Xin and Li, Yu , journal=. Self-supervised contrastive learning for integrative single cell. 2022 , m...

    2022

  23. [31]

    Nature Machine Intelligence , volume=

    Contrastive learning enables rapid mapping to multimodal single-cell atlas of multimillion scale , author=. Nature Machine Intelligence , volume=. 2022 , doi=

    2022

  24. [32]

    Advances in Neural Information Processing Systems , volume=

    Tabula: A Tabular Self-Supervised Foundation Model for Single-Cell Transcriptomics , author=. Advances in Neural Information Processing Systems , volume=. 2025 , url=

    2025

  25. [33]

    Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing , month=nov, year=

    Gao, Tianyu and Yao, Xingcheng and Chen, Danqi , editor=. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing , month=nov, year=. doi:10.18653/v1/2021.emnlp-main.552 , url=

    2021 doi

  26. [34]

    Aurel and Tabrizi, Shervin and Hodis, Eran and Melton, Douglas A

    Kotliar, Dylan and Veres, Adrian and Nagy, M. Aurel and Tabrizi, Shervin and Hodis, Eran and Melton, Douglas A. and Sabeti, Pardis C. , journal=. Identifying gene expression programs of cell-type identity and cellular activity with single-cell. 2019 , doi=

    2019

  27. [35]

    Advances in Neural Information Processing Systems , volume=

    Can contrastive learning avoid shortcut solutions? , author=. Advances in Neural Information Processing Systems , volume=. 2021 , url=

    2021

  28. [36]

    , journal=

    Ji, Yanrong and Zhou, Zhihan and Liu, Han and Davuluri, Ramana V. , journal=. 2021 , doi=

    2021

  29. [37]

    Nature Methods , volume=

    Nucleotide Transformer: building and evaluating robust foundation models for human genomics , author=. Nature Methods , volume=. 2025 , doi=

    2025

  30. [38]

    2025 , doi=

    Long, Weicai and Su, Houcheng and Xiong, Jiaqi and Zhang, Yanlin , journal=. 2025 , doi=

    2025

  31. [39]

    Interpretable

    Chen, Jiayang and Hu, Zhihang and Sun, Siqi and Tan, Qingxiong and Wang, Yixuan and Yu, Qinze and Zong, Licheng and Hong, Liang and Xiao, Jin and Shen, Tao and King, Irwin and Li, Yu , year=. Interpretable. 2204.00300 , archivePrefix=

    arXiv

  32. [40]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    Iterative Closed-Loop Motion Synthesis for Scaling the Capabilities of Humanoid Control , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=. 2602.21599 , archivePrefix=

  33. [41]

    Nature Methods , volume=

    Benchmarking algorithms for gene regulatory network inference from single-cell transcriptomic data , author=. Nature Methods , volume=. 2020 , doi=

    2020

  34. [42]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

  35. [43]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

  36. [44]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    1980

  37. [45]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

  38. [46]

    Clancey and Glenn Rennels , abstract =

    Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

    1984 doi

  39. [47]

    and Rennels, Glenn R

    Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

  40. [48]

    Poligon: A System for Parallel Problem Solving

    Rice, James. Poligon: A System for Parallel Problem Solving

  41. [49]

    Transfer of Rule-Based Expertise through a Tutorial Dialogue

    Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

  42. [50]

    The Engineering of Qualitative Models

    Clancey, William J. The Engineering of Qualitative Models

  43. [51]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  44. [52]

    Pluto: The 'Other' Red Planet

    NASA. Pluto: The 'Other' Red Planet

Pith tools

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