CytoSAE: Interpretable Cell Embeddings for Hematology

Carsten Marr; Hyesu Lim; Katharina S. G\"otze; Michele Buck; Muhammed Furkan Dasdelen; Steffen Schneider

REVIEW 4 major objections 4 minor 29 references

CytoSAE: Interpretable Cell Embeddings for Hematology

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 CytoSAE shows that sparse autoencoders can turn hematology embeddings into transferable morphological concepts and explainable AML-subtype classifiers.

desk verdict A genuinely new application of SAEs to hematology with a credible qualitative core, but the AML classification F1 needs baseline and split controls before it carries weight. read the letter →

arxiv 2507.12464 v1 pith:W3L7WCIK submitted 2025-07-16 cs.CV cs.LGq-bio.QM

Muhammed Furkan Dasdelen , Hyesu Lim , Michele Buck , Katharina S. Götze , Carsten Marr , Steffen Schneider This is my paper

classification cs.CVcs.LGq-bio.QM

keywords sparseautoencoderhematologycellmorphologyinterpretabilityacutemyeloidleukemiaconceptdiscoveryfoundationmodelspatch-levelattribution

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 sets out to show that sparse autoencoders, a dictionary-learning tool from interpretability research, can make a general-purpose cell-image foundation model explainable at the sub-cellular level. The authors train CytoSAE on the internal per-patch vectors of DinoBloom-B, a hematology foundation model, using 41,906 peripheral-blood single-cell images and expanding each 768-dimensional vector into 49,152 sparse latent units. They argue that the learned units are genuine morphological concepts because the same units fire on matching structures across four external datasets, including bone marrow cytology, and because an expert cytomorphologist confirms the recovered patterns. Counting active units per patient produces barcodes on which logistic regression reaches a weighted F1 of $0.832 \pm 0.044$ for four genetic AML subtypes and healthy controls, comparable to the full deep-learning baseline. The contribution is a concrete route from black-box cell embeddings to disease-specific, visually checkable evidence.

What carries the argument

The carrying mechanism is a sparse autoencoder with a 64-fold expansion: a 768-dimensional token $x$ from the backbone's residual stream is encoded as $z = \mathrm{ReLU}(W_{\mathrm{enc}}(x - b_{\mathrm{dec}}) + b_{\mathrm{enc}}) \in \mathbb{R}^{49{,}152}$ and decoded as $\hat{x} = W_{\mathrm{dec}} z + b_{\mathrm{dec}}$, trained by MSE reconstruction plus an L1 penalty on $z$, with the decoder bias initialized at the geometric median of training tokens and ghost-gradient resampling to revive dead latents. Each latent index $s$ is treated as one morphological concept. A threshold $\tau$ binarizes patch-level activations, counting active patches to form image-, patient-, and disease-level barcodes; multiplying a patch by its activation yields a concept-attribution map. This hierarchical counting is what converts an unsupervised dictionary into a patient-level diagnostic feature space.

What would settle it

One decisive test would be to take a latent that fires on eosinophilic granules in peripheral blood and check whether the same latent index fires on eosinophilic granules in bone marrow images from a different institution, scanner, and stain protocol; if the mapping breaks, the cross-dataset consistency is an acquisition artifact rather than shared morphology. A second check is to train the identical SAE on a re-initialized DinoBloom-B backbone and see whether the same concepts reappear.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that CytoSAE disentangles the compact embedding of a hematology foundation model into a reusable dictionary of morphological concepts, and that these concepts carry the diagnostic signal. The model decomposes the layer-11 residual stream of DinoBloom-B with an expansion factor of 64, producing 49,152 latents; at a mean-activation threshold of $-3$ on a log scale, roughly 2,936 concepts are confidently activated. The same latents generalize to peripheral blood datasets and to 171,373 expert-annotated bone marrow cells, and patch-level activation maps localize structures such as eosinophilic granules consistently across datasets. When patient-level barcodes are classified, the weighted F1 of $0.832 \pm 0.044$ is comparable to the deep-learning baseline, and differential disease barcodes identify subtype-specific morphologies such as vacuolated monocytic precursors for CBFB::MYH11 and granulated blasts for PML::RARA.

Load-bearing premise

CytoSAE is only a linear dictionary on top of DinoBloom-B's layer-11 residual stream, so everything depends on those embeddings already storing cell morphology in a stable, sparse, linearly decodable form; if they mostly store staining or scanner style, the cross-dataset consistency would be an artifact of shared acquisition rather than shared biology.

Editorial extensions

If this is right

A single SAE trained on peripheral blood can be applied directly to bone marrow cytology, so explainable concept extraction does not require retraining per specimen type.
Patch-level attribution makes the basis of a patient-level prediction inspectable, letting clinicians verify that the model is looking at structures such as eosinophilic granules rather than at global slide artifacts.
Keeping only confidently activated concepts improves classification stability, which suggests that pruning the dictionary to meaningful latents is safe for downstream tasks.
Disease-level barcodes provide a data-driven shortlist of morphological hallmarks for a genetic AML subtype, which can then be checked by expert review.

Reading between the lines

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

Beyond the paper, if the cross-dataset consistency is biological rather than acquisition-driven, the same dictionary could serve as a fixed morphological atlas for rare hematologic diseases, but the paper does not test rare-disease coverage.
A natural extension would be applying the same SAE recipe to residual streams of other medical foundation models; the paper only tests one backbone.
Because only about 2,936 of 49,152 latents are confidently activated, most of the dictionary capacity is idle; pruning to the active subset could make the method cheaper at deployment, though the paper does not evaluate this.
The barcode representation is linear in concept counts, so it could be plugged into survival or regression models with clinical covariates, while the paper demonstrates classification only.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

CytoSAE is the first sparse autoencoder applied to hematology cell images, and the qualitative core holds up. The authors train a 49,152-unit SAE on DinoBloom-B residual tokens from ~42k peripheral blood cells, then show the discovered concepts transfer to four other datasets including bone marrow, with patch-level attribution that is visually convincing. The patient- and disease-level barcode aggregation is a new analytical step for this domain, and they ship code and weights, which makes the whole thing reproducible. The eosinophil granule example is a clean demonstration of sub-cellular attribution.

The soft spots are real but not disqualifying. Expert validation is thin: 50 latents sampled via k-means, and in the disease-specific comparison only 10 of 50 PML::RARA concepts were judged meaningful. That is reported honestly, but it does not support a strong claim about disease-specific morphology for that subtype. The AML classification result (weighted F1 0.832±0.044) is not yet attributable to CytoSAE. There is no patient-stratified split description, no raw-embedding linear-probe baseline, and the activation threshold -3 was chosen by varying the number of latents and observing performance on the same data. With 189 patients and ~2,936 features, selection bias and possible leakage are real concerns. The stress-test note is on target; I don't think the authors cheated, but the F1 needs controls before it can be compared to the deep-learning baseline. Additionally, the claim that performance remains stable until meaningful concepts are removed is confounded by threshold selection on the same data.

The central qualitative claim—that a SAE on residual-stream features discovers stable, transferable morphological concepts—holds up. This is a legitimate application of an established method to a new domain, with a thoughtful evaluation structure and honest limitation statements. The citation pattern is fine; the backbone and training data come from the same group, but that is disclosed and reasonable for a methods-application paper. It deserves peer review. The referee should ask for: (1) split/fold details for the barcode classifier and a raw DinoBloom-B embedding baseline; (2) held-out threshold selection or nested validation; (3) more quantitative expert validation, particularly for PML::RARA. Those are fixable.

Who is this for? Anyone working on interpretability for medical imaging, especially hematology foundation models. It is not a foundational methods paper, but it is a solid domain application with reproducible artifacts. Recommendation: accept for peer review with the expectation of a revision.

Referee Report

4 major / 4 minor

Summary. The paper proposes CytoSAE, a sparse autoencoder trained on DinoBloom-B residual-stream tokens extracted from 41,906 peripheral blood single-cell images. The model expands 768-dimensional tokens to a 49,152-dimensional sparse latent space using an MSE reconstruction loss with an L1 penalty, then binarizes patch-level activations and aggregates them into image-, patient-, and disease-level barcodes. The authors report that the discovered latents correspond to morphologically meaningful concepts, that these concepts transfer across four external datasets including bone marrow cytology, and that logistic regression on patient barcodes classifies four AML subtypes plus a healthy control group with weighted F1 0.832±0.044, comparable to a fully deep learning baseline, while offering sub-cellular patch-level explanations.

Significance. If the claims hold, CytoSAE is a valuable interpretability layer for hematology foundation models: it provides a structured, patch-attributable vocabulary of morphological concepts and a transparent patient-level representation for subtype classification. The paper's strengths include the release of source code and model weights, the use of established public datasets, the involvement of an expert cytomorphologist, and explicit ablation studies over expansion factors, sparsity coefficients, residual layers, and ghost gradient resampling. The main caveat is that the quantitative claims, especially the classification result and the disease-specific concept validation, need stronger evaluation protocols before the conclusions can be fully supported.

major comments (4)

[Sec. 3.3, Eq. (3), Fig. 4] The key quantitative claim, weighted F1 0.832±0.044 for AML subtype classification from patient barcodes, lacks essential evaluation detail. The paper does not state whether classification used patient-stratified k-fold cross-validation or a fixed split, how many folds or runs produced the ±0.044 interval, or how the 189 patients were partitioned. Because barcodes aggregate multiple images per patient, any non-patient-stratified split can leak information between training and test sets. In addition, no linear-probe baseline on raw DinoBloom-B embeddings is reported; without such a baseline on, for example, patient-level mean CLS or patch tokens under the identical protocol, the result cannot be attributed to CytoSAE concepts rather than to the backbone features. Please report the exact evaluation protocol and add this baseline.
[Sec. 3.3 and Fig. 4A] The activation threshold tau in Eq. (2) is selected by varying the number of latents and monitoring classification performance on the same data. With 189 patients and roughly 2,936 features at the chosen threshold, selecting the threshold on the evaluation data can optimistically bias the reported F1. The threshold selection procedure should be described on a validation split or through nested cross-validation, and performance should be reported across the full threshold range with error bars, not only at the selected threshold of -3.
[Sec. 3.3, disease-barcode paragraph] The expert validation of disease-specific concepts is ambiguous and weak for PML::RARA. The text reports "5 out of 10 and 32 out of 50" for CBFB::MYH11-specific concepts and "4 out of 10 and 10 out of 50" for PML::RARA-specific concepts, but the sample sizes and the relationship between the two numbers are unclear. As written, only 10 out of 50 of the top PML::RARA features were judged disease-specific, which substantially weakens the claim that disease-specific concepts are discovered. Please clarify the full contingency table, report how the 10 and 50 subsets were constructed, and compare against chance agreement.
[Sec. 3.1, Fig. 2] The cross-dataset generalization claim is supported mainly by qualitative visualization of selected latents and images. No quantitative measure of concept consistency across datasets is provided, so it is difficult to distinguish shared morphology from backbone style or batch effects. A quantitative evaluation, such as measuring how often the top activating patches for a latent in MLL23 are also top activations in Acevedo, Matek19, or BMC, or comparing activation statistics across datasets, would make the transfer claim more robust.

minor comments (4)

[Fig. 1B caption] The caption reads "within the patent's single-cell images" and should read "within the patient's single-cell images."
[Sec. 3.4] There are typos: "influced" should be "influenced" and "treshold" should be "threshold."
[Sec. 2, Eq. (2)] The threshold tau is used in Eq. (2) but never explicitly defined; the text later refers to thresholds in log10 mean activation units. Please define tau in the text or state that it is the mean activation threshold.
[Sec. 3.3] The comparison to the deep learning baseline [12] is reported only as "similar" without a statistical comparison; please provide confidence intervals or a significance test for the difference in weighted F1.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; SAE training is unsupervised and classification is a separate linear probe.

full rationale

The derivation chain is not circular. CytoSAE is trained with Eq. (1), an MSE reconstruction loss plus L1 sparsity penalty, on unlabeled DinoBloom-B residual-stream tokens from MLL23; disease labels never enter the SAE objective. Patient barcodes ap are defined in Eq. (3) as normalized counts of thresholded latent activations, and AML subtype classification is a separate logistic-regression probe on those barcodes. There is no equation in which the classification target defines the SAE latents or the binarization threshold, and no fitted parameter is renamed as a prediction. The self-citations (DinoBloom-B [14], MLL23 [25], AML-Hehr [12], Matek19 [21], BMC [20]) are provenance overlaps with the authors' group, but each cited artifact is an externally falsifiable model or dataset, and the SAE's concept discovery is evaluated post hoc by expert annotation and cross-dataset retrieval rather than by construction. The main caveats are evaluation-protocol concerns rather than circularity: the paper does not specify whether the AML classification used patient-stratified cross-validation, provides no linear-probe baseline on raw DinoBloom-B embeddings, and selects the -3 latent threshold after inspecting performance on the same data. These could make the reported F1 optimistic, but they do not make the result equivalent to the input by definition.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

The central claim rests on the backbone representation, the SAE monosemanticity assumption, and counting-based aggregation. The hand-set numbers are the expansion factor, L1 coefficient, and activation threshold. No physical entities are introduced; the SAE latents are internal model components with no independent falsifiable handle beyond expert-labeled images.

free parameters (3)

Expansion factor d_SAE / d_m = 64
Selected by ablation on MLL23; higher expansion yields more latents above threshold but similar high-activation cluster size.
L1 coefficient lambda = 8e-5
Selected by ablation; 10x higher collapses concepts to 128±83, 10x lower inflates L0 by two orders of magnitude.
Activation binarization threshold tau = -3 (log10 scale)
Used for image/patient barcodes and for the reported F1=0.832; a separate held-out threshold selection procedure is not described.

assumptions (3)

domain assumption DinoBloom-B layer-11 residual stream embeddings linearly encode the cytomorphological information needed for hematological diagnosis.
CytoSAE decomposes only this representation; if the backbone does not separate morphology from staining/background style, the discovered concepts will not transfer.
domain assumption A single L1-regularized linear dictionary on raw residual stream tokens yields monosemantic, interpretable concepts.
This is the core SAE assumption imported from the mechanistic-interpretability literature; the paper does not test monosemanticity directly beyond visual inspection.
domain assumption Patch-level latent activations can be binarized at a fixed threshold and counted to produce meaningful image-, patient-, and disease-level summaries.
Equations 2 and 3 assume counting activated patches is a faithful summary; no ablations on the counting scheme are reported.

invented entities (1)

SAE latent concepts (dictionary atoms)
purpose: To serve as interpretable morphological building blocks that aggregate into patient and disease barcodes.
The latents are internal model parameters; the only external validation is qualitative expert inspection of reference images, with no quantitative, independently repeatable handle such as a predicted measurement.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CytoSAE: Interpretable Cell Embeddings for Hematology." pith.science (2026). https://pith.science/paper/W3L7WCIK

@misc{pith2026250712464,
  author       = {Pith},
  title        = {Pith review of: CytoSAE: Interpretable Cell Embeddings for Hematology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W3L7WCIK}},
  note         = {Machine review of arXiv:2507.12464}
}

read the original abstract

Sparse autoencoders (SAEs) emerged as a promising tool for mechanistic interpretability of transformer-based foundation models. Very recently, SAEs were also adopted for the visual domain, enabling the discovery of visual concepts and their patch-wise attribution to tokens in the transformer model. While a growing number of foundation models emerged for medical imaging, tools for explaining their inferences are still lacking. In this work, we show the applicability of SAEs for hematology. We propose CytoSAE, a sparse autoencoder which is trained on over 40,000 peripheral blood single-cell images. CytoSAE generalizes to diverse and out-of-domain datasets, including bone marrow cytology, where it identifies morphologically relevant concepts which we validated with medical experts. Furthermore, we demonstrate scenarios in which CytoSAE can generate patient-specific and disease-specific concepts, enabling the detection of pathognomonic cells and localized cellular abnormalities at the patch level. We quantified the effect of concepts on a patient-level AML subtype classification task and show that CytoSAE concepts reach performance comparable to the state-of-the-art, while offering explainability on the sub-cellular level. Source code and model weights are available at https://github.com/dynamical-inference/cytosae.

Figures

Figures reproduced from arXiv: 2507.12464 by the authors.

**Figure 1.** Overview. (A) CytoSAE is trained on over 40K peripheral blood singlecell images using embeddings from DinoBloom-B, disentangling compact embedding information into meaningful concepts. We employ MSE loss for embedding reconstruction and apply L1 normalization to SAE latents. (B) Patient-level information is aggregated by counting the number of activations for each concept within the patent’s single-cell images, f… view at source ↗

**Figure 2.** CytoSAE discovers morphologically relevant concepts across [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Patch-level morphological components are consistent across di [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Patient-wise concepts can predict disease. (A) [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Disease-wise comparison reveals disease-specific morphological [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 12 canonical work pages

[1]

arXiv preprint arXiv:2410.03334 (2024)

Abdulaal, A., Fry, H., Montaña-Brown, N., Ijishakin, A., Gao, J., Hyland, S., Alexander, D.C., Castro, D.C.: An x-ray is worth 15 features: Sparse autoencoders for interpretable radiology report generation. arXiv preprint arXiv:2410.03334 (2024)

arXiv 2024
[2]

arXiv preprint arXiv:2005.00928 (2020)

Abnar, S., Zuidema, W.: Quantifying attention flow in transformers. arXiv preprint arXiv:2005.00928 (2020)

arXiv 2020
[3]

Data in brief30, 105474 (2020)

Acevedo, A., Merino, A., Alférez, S., Molina, Á., Boldú, L., Rodellar, J.: A dataset of microscopic peripheral blood cell images for development of automatic recogni- tion systems. Data in brief30, 105474 (2020)

2020
[4]

Transformer Circuits Thread (2023), https://transformer-circuits.pub/2023/monosemantic- features/index.html

Bricken, T., Templeton, A., Batson, J., Chen, B., Jermyn, A., Conerly, T., Turner, N., Anil, C., Denison, C., Askell, A., Lasenby, R., Wu, Y., Kravec, S., Schiefer, N., Maxwell, T., Joseph, N., Hatfield-Dodds, Z., Tamkin, A., Nguyen, K., McLean, B., Burke, J.E., Hume, T., Carter, S., Henighan, T., Olah, C.: Towards monose- manticity: Decomposing language ...

2023
[5]

NPJ digital medicine5(1), 156 (2022)

Chen, H., Gomez, C., Huang, C.M., Unberath, M.: Explainable medical imaging ai needs human-centered design: guidelines and evidence from a systematic review. NPJ digital medicine5(1), 156 (2022)

work page 2022
[6]

Nature Medicine30(3), 850–862 (2024)

Chen, R.J., Ding, T., Lu, M.Y., Williamson, D.F., Jaume, G., Song, A.H., Chen, B., Zhang, A., Shao, D., Shaban, M., et al.: Towards a general-purpose foundation model for computational pathology. Nature Medicine30(3), 850–862 (2024)

2024
[7]

arXiv preprint arXiv:2501.18052 (2025) 10 Dasdelen et al

Cywiński,B.,Deja,K.:Saeuron:Interpretableconceptunlearningindiffusionmod- els with sparse autoencoders. arXiv preprint arXiv:2501.18052 (2025) 10 Dasdelen et al

arXiv 2025
[8]

Transformer Circuits Thread (2021), https://transformer-circuits.pub/2021/framework/index.html

Elhage, N., Nanda, N., Olsson, C., Henighan, T., Joseph, N., Mann, B., Askell, A., Bai, Y., Chen, A., Conerly, T., DasSarma, N., Drain, D., Ganguli, D., Hatfield- Dodds, Z., Hernandez, D., Jones, A., Kernion, J., Lovitt, L., Ndousse, K., Amodei, D., Brown, T., Clark, J., Kaplan, J., McCandlish, S., Olah, C.: A mathemat- ical framework for transformer circ...

2021

Show all 29 references

[9]

European Parliament and Council: in Regulation of the European Parliament and of the Council laying down harmonised rules on Artificial Intelligence (Artificial Intelligence Act) (2024), (ed European Parliament and Council)

2024
[10]

Fry, H.: Towards multimodal interpretability: Learning sparse interpretable fea- tures in vision transformers (Apr 30 2024), https://www.lesswrong.com/posts/ iYFuZo9BMvr6GgMs5/case-study-interpreting-manipulating-and-controlling-clip

2024
[11]

arXiv preprint arXiv:2406.04093 (2024)

Gao, L., la Tour, T.D., Tillman, H., Goh, G., Troll, R., Radford, A., Sutskever, I., Leike, J., Wu, J.: Scaling and evaluating sparse autoencoders. arXiv preprint arXiv:2406.04093 (2024)

2024 arXiv
[12]

PLOS Digital Health2(3), e0000187 (2023)

Hehr, M., Sadafi, A., Matek, C., Lienemann, P., Pohlkamp, C., Haferlach, T., Spiekermann, K., Marr, C.: Explainable ai identifies diagnostic cells of genetic aml subtypes. PLOS Digital Health2(3), e0000187 (2023)

2023
[13]

In: The Twelfth In- ternational Conference on Learning Representations (2023)

Huben, R., Cunningham, H., Smith, L.R., Ewart, A., Sharkey, L.: Sparse autoen- coders find highly interpretable features in language models. In: The Twelfth In- ternational Conference on Learning Representations (2023)

2023
[14]

In: International Conference on Medical Image Computing and Computer-Assisted Intervention

Koch, V., Wagner, S.J., Kazeminia, S., Sancar, E., Hehr, M., Schnabel, J.A., Peng, T., Marr, C.: Dinobloom: a foundation model for generalizable cell embeddings in hematology. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. pp. 520–5...

2024
[15]

In: Advancements In Medical Foundation Models: Explain- ability, Robustness, Security, and Beyond (2024)

Le, N.M., Patel, N., Shen, C., Martin, B., Eng, A., Shah, C., Grullon, S., Juyal, D.: Learning biologically relevant features in a pathology foundation model using sparse autoencoders. In: Advancements In Medical Foundation Models: Explain- ability, Robustness, Security, and B...

2024
[16]

arXiv preprint arXiv:2408.05147 (2024)

Lieberum, T., Rajamanoharan, S., Conmy, A., Smith, L., Sonnerat, N., Varma, V., Kramár, J., Dragan, A., Shah, R., Nanda, N.: Gemma scope: Open sparse autoencoders everywhere all at once on gemma 2. arXiv preprint arXiv:2408.05147 (2024)

2024 arXiv
[17]

arXiv preprint arXiv:2412.05276 (2024)

Lim, H., Choi, J., Choo, J., Schneider, S.: Sparse autoencoders reveal selective remapping of visual concepts during adaptation. arXiv preprint arXiv:2412.05276 (2024)

2024 arXiv
[18]

arXiv preprint arXiv:1312.5663 (2013)

Makhzani, A., Frey, B.: K-sparse autoencoders. arXiv preprint arXiv:1312.5663 (2013)

2013 arXiv
[19]

arXiv preprint arXiv:2403.19647 (2024)

Marks, S., Rager, C., Michaud, E.J., Belinkov, Y., Bau, D., Mueller, A.: Sparse feature circuits: Discovering and editing interpretable causal graphs in language models. arXiv preprint arXiv:2403.19647 (2024)

2024 arXiv
[20]

https://doi.org/10.7937/TCIA.AXH3-T579, data set

Matek, C., Krappe, S., Münzenmayer, C., Haferlach, T., Marr, C.: An expert- annotated dataset of bone marrow cytology in hematologic malignancies (2021). https://doi.org/10.7937/TCIA.AXH3-T579, data set

2021 doi
[21]

https: //doi.org/10.7937/tcia.2019.36f5o9ld, data set

Matek, C., Schwarz, S., Marr, C., Spiekermann, K.: A single-cell morphological dataset of leukocytes from aml patients and non-malignant controls (2019). https: //doi.org/10.7937/tcia.2019.36f5o9ld, data set

2019 doi
[22]

Na- ture616(7956), 259–265 (2023) CytoSAE: Interpretable Cell Embeddings for Hematology 11

Moor, M., Banerjee, O., Abad, Z.S.H., Krumholz, H.M., Leskovec, J., Topol, E.J., Rajpurkar, P.: Foundation models for generalist medical artificial intelligence. Na- ture616(7956), 259–265 (2023) CytoSAE: Interpretable Cell Embeddings for Hematology 11

2023
[23]

arXiv preprint arXiv:2404.16014 (2024)

Rajamanoharan, S., Conmy, A., Smith, L., Lieberum, T., Varma, V., Kramár, J., Shah, R., Nanda, N.: Improving dictionary learning with gated sparse autoen- coders. arXiv preprint arXiv:2404.16014 (2024)

2024 arXiv
[24]

arXiv preprint arXiv:2501.16496 (2025)

Sharkey, L., Chughtai, B., Batson, J., Lindsey, J., Wu, J., Bushnaq, L., Goldowsky- Dill, N., Heimersheim, S., Ortega, A., Bloom, J., et al.: Open problems in mecha- nistic interpretability. arXiv preprint arXiv:2501.16496 (2025)

2025 arXiv
[25]

medRxiv pp

Shetab Boushehri, S., Gruber, A., Kazeminia, S., Matek, c., Spiekermann, K., Pohlkamp, C., Haferlach, T., Marr, C.: A large expert-annotated single-cell pe- ripheral blood dataset for hematological disease diagnostics. medRxiv pp. 2025–02 (2025)

2025
[26]

arXiv preprint arXiv:2502.06755 (2025)

Stevens, S., Chao, W.L., Berger-Wolf, T., Su, Y.: Sparse autoencoders for scien- tifically rigorous interpretation of vision models. arXiv preprint arXiv:2502.06755 (2025)

2025
[27]

Anthropic (2024)

Templeton, A.: Scaling monosemanticity: Extracting interpretable features from claude 3 sonnet. Anthropic (2024)

2024
[28]

Cancer Cell41(9), 1650–1661 (2023)

Wagner, S.J., Reisenbüchler, D., West, N.P., Niehues, J.M., Zhu, J., Foersch, S., Veldhuizen, G.P., Quirke, P., Grabsch, H.I., van den Brandt, P.A., et al.: Transformer-based biomarker prediction from colorectal cancer histology: A large- scale multicentric study. Cancer Cell4...

2023
[29]

In: Proceedings of the IEEE conference on computer vision and pattern recognition

Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., Torralba, A.: Learning deep features for discriminative localization. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 2921–2929 (2016)

2016

Pith tools

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

[1] [1]

arXiv preprint arXiv:2410.03334 (2024)

Abdulaal, A., Fry, H., Montaña-Brown, N., Ijishakin, A., Gao, J., Hyland, S., Alexander, D.C., Castro, D.C.: An x-ray is worth 15 features: Sparse autoencoders for interpretable radiology report generation. arXiv preprint arXiv:2410.03334 (2024)

arXiv 2024

[2] [2]

arXiv preprint arXiv:2005.00928 (2020)

Abnar, S., Zuidema, W.: Quantifying attention flow in transformers. arXiv preprint arXiv:2005.00928 (2020)

arXiv 2020

[3] [3]

Data in brief30, 105474 (2020)

Acevedo, A., Merino, A., Alférez, S., Molina, Á., Boldú, L., Rodellar, J.: A dataset of microscopic peripheral blood cell images for development of automatic recogni- tion systems. Data in brief30, 105474 (2020)

2020

[4] [4]

Transformer Circuits Thread (2023), https://transformer-circuits.pub/2023/monosemantic- features/index.html

Bricken, T., Templeton, A., Batson, J., Chen, B., Jermyn, A., Conerly, T., Turner, N., Anil, C., Denison, C., Askell, A., Lasenby, R., Wu, Y., Kravec, S., Schiefer, N., Maxwell, T., Joseph, N., Hatfield-Dodds, Z., Tamkin, A., Nguyen, K., McLean, B., Burke, J.E., Hume, T., Carter, S., Henighan, T., Olah, C.: Towards monose- manticity: Decomposing language ...

2023

[5] [5]

NPJ digital medicine5(1), 156 (2022)

Chen, H., Gomez, C., Huang, C.M., Unberath, M.: Explainable medical imaging ai needs human-centered design: guidelines and evidence from a systematic review. NPJ digital medicine5(1), 156 (2022)

work page 2022

[6] [6]

Nature Medicine30(3), 850–862 (2024)

Chen, R.J., Ding, T., Lu, M.Y., Williamson, D.F., Jaume, G., Song, A.H., Chen, B., Zhang, A., Shao, D., Shaban, M., et al.: Towards a general-purpose foundation model for computational pathology. Nature Medicine30(3), 850–862 (2024)

2024

[7] [7]

arXiv preprint arXiv:2501.18052 (2025) 10 Dasdelen et al

Cywiński,B.,Deja,K.:Saeuron:Interpretableconceptunlearningindiffusionmod- els with sparse autoencoders. arXiv preprint arXiv:2501.18052 (2025) 10 Dasdelen et al

arXiv 2025

[8] [8]

Transformer Circuits Thread (2021), https://transformer-circuits.pub/2021/framework/index.html

Elhage, N., Nanda, N., Olsson, C., Henighan, T., Joseph, N., Mann, B., Askell, A., Bai, Y., Chen, A., Conerly, T., DasSarma, N., Drain, D., Ganguli, D., Hatfield- Dodds, Z., Hernandez, D., Jones, A., Kernion, J., Lovitt, L., Ndousse, K., Amodei, D., Brown, T., Clark, J., Kaplan, J., McCandlish, S., Olah, C.: A mathemat- ical framework for transformer circ...

2021