mlsanityData to 14 Oct 2026

Interpreting biomedical and science models

60 papers0 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: protein, medical, features, interpretability, sparse

Key facts. Interpreting biomedical and science models: 60 papers on arXiv since 2024 (data through 14 Oct 2026). 17 of its 60 mapped papers (28%) cite “InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoders” (2024, arXiv 2412.12101). 1 new paper 24 Sep–14 Oct.

What’s new

24 Sep – 14 Oct 2026

1 new since 24 Sep

  1. Generative modeling of intrinsically disordered protein regions by reinforcing sparse autoencoder features

Foundations

Most cited by papers here

  1. InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoders2024cited by 17 here
  2. Interpreting and Steering Protein Language Models through Sparse Autoencoders2025cited by 9 here
  3. Towards Interpretable Protein Structure Prediction with Sparse Autoencoders2025cited by 8 here
  4. Learning biologically relevant features in a pathology foundation model using sparse autoencoders2024cited by 7 here
  5. An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generation2024cited by 7 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2022Toy Models of Superpositioncited by 9 here
  2. 2023Towards Automated Circuit Discovery for Mechanistic Interpretabilitycited by 5 here
  3. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 26 here
  4. 2024Improving Dictionary Learning with Gated Sparse Autoencoderscited by 5 here
  5. 2024Scaling and evaluating sparse autoencoderscited by 20 here
  6. 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 6 here
  7. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 6 here
  8. 2024A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoderscited by 5 here
  9. 2025Learning Multi-Level Features with Matryoshka Sparse Autoencoderscited by 8 here
15 more it builds on, not in this chain
  1. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 7 here
  2. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 4 here
  3. 2024Mechanistic Interpretability for AI Safety -- A Reviewcited by 6 here
  4. 2024Transcoders Find Interpretable LLM Feature Circuitscited by 4 here
  5. 2024Learning biologically relevant features in a pathology foundation model using sparse autoencoderscited by 7 here
  6. 2024An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generationcited by 7 here
  7. 2024Automatically Interpreting Millions of Features in Large Language Modelscited by 4 here
  8. 2024InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoderscited by 17 here
  9. 2024Sparse autoencoders reveal selective remapping of visual concepts during adaptationcited by 3 here
  10. 2024BatchTopK Sparse Autoencoderscited by 6 here
  11. 2025Sparse Autoencoders Trained on the Same Data Learn Different Featurescited by 5 here
  12. 2025Interpretable and Testable Vision Features via Sparse Autoencoderscited by 4 here
  13. 2025Interpreting and Steering Protein Language Models through Sparse Autoencoderscited by 9 here
  14. 2025Towards Interpretable Protein Structure Prediction with Sparse Autoencoderscited by 8 here
  15. 2026Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnetcited by 5 here

The 24 most cited of 92.

Browse

30 of 60, ranked 6–35

  1. 6Systematic Evaluation of Single-Cell Foundation Model Interpretability Reveals Attention Captures Co-Expression Rather Than Unique Regulatory Signal20263
  2. 7CytoSAE: Interpretable Cell Embeddings for Hematology20253
  3. 8Sparse Autoencoders for Low-N Protein Function Prediction and Design20252
  4. 9MedSAE: Dissecting MedCLIP Representations with Sparse Autoencoders20252
  5. 10Protein Circuit Tracing via Cross-layer Transcoders20261
  6. 11Suppressing Prior-Comparison Hallucinations in Radiology Report Generation via Semantically Decoupled Latent Steering20261
  7. 12Sparse autoencoders reveal organized biological knowledge but minimal regulatory logic in single-cell foundation models: a comparative atlas of Geneformer and scGPT20261
  8. 13Automated Neuron Labelling Enables Generative Steering and Interpretability in Protein Language Models20251
  9. 14Mammo-SAE: Interpreting Breast Cancer Concept Learning with Sparse Autoencoders20251
  10. 15ProtSAE: Disentangling and Interpreting Protein Language Models via Semantically-Guided Sparse Autoencoders20251
All 30
  1. 16DILA: Dictionary Label Attention for Mechanistic Interpretability in High-dimensional Multi-label Medical Coding Prediction20241
  2. 17Can sparse autoencoders make sense of gene expression latent variable models?20241
  3. 18Beyond Label Attention: Transparency in Language Models for Automated Medical Coding via Dictionary Learning20241
  4. 19MATEX: Multi-scale Attention and Text-guided Explainability of Medical Vision-Language Models20260
  5. 20A Monosemantic Attribution Framework for Stable Interpretability in Clinical Neuroscience Transformer-Based Language Models20260
  6. 21The Patient is not a Moving Document: A World Model Training Paradigm for Longitudinal EHR20260
  7. 22Causal Circuit Tracing Reveals Distinct Computational Architectures in Single-Cell Foundation Models: Inhibitory Dominance, Biological Coherence, and Cross-Model Convergence20260
  8. 23Information Routing in Atomistic Foundation Models: How Task Alignment and Equivariance Shape Linear Disentanglement20260
  9. 24Tracing Pharmacological Knowledge In Large Language Models20260
  10. 25Exhaustive Circuit Mapping of a Single-Cell Foundation Model Reveals Massive Redundancy, Heavy-Tailed Hub Architecture, and Layer-Dependent Differentiation Control20260
  11. 26Interpretability without actionability: mechanistic methods cannot correct language model errors despite near-perfect internal representations20260
  12. 27Sparse Autoencoders for Interpretable Medical Image Representation Learning20260
  13. 28Sparse Autoencoder Decomposition of Clinical Sequence Model Representations: Feature Complexity, Task Specialisation, and Mortality Prediction20260
  14. 29GeoSAE: Geometric Prior-Guided Layer-Wise Sparse Autoencoder Annotation of Brain MRI Foundation Models20260
  15. 30Molecules Meet Language: Confound-Aware Representation Learning and Chemical Property Steering in Transformer-VAE Latent Spaces20260
  16. 31Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders20260
  17. 32Universal Boosts, Specific Suppressors: Sparse Autoencoder Steering of Medical Vision-Language Models20260
  18. 33VFUSE: Virulent Feature Understanding with Sparse autoEncoders20260
  19. 34Circuit Tracing in Autoregressive Protein Language Models20260
  20. 35Translating Inference-Time Control to Radiology Vision-Language Models: Activation Steering for Pneumonia Classification on Chest X-rays20260