Mechanistic Interpretability
Reverse-engineering neural networks: superposition, sparse autoencoders and transcoders, circuits and attribution graphs, probing and activation steering.
Most influential
- Systems biology and gene networks in neurodevelopmental and neurodegenerative disordersReview447 citationsdoi:10.1038/nrg3934
- Exploring single-cell data with deep multitasking neural networksArticle297 citationsdoi:10.1038/s41592-019-0576-7
- Decoding of position in the developing neural tube from antiparallel morphogen gradientsArticle184 citationsdoi:10.1126/science.aam5887
- Biologically-informed neural networks guide mechanistic modeling from sparse experimental data
Biologically-informed neural networks (BINNs), an extension of physics-informed neural networks [1], are introduced and used to discover the underlying dynamics of biological systems from sparse experimental data. In the present work, BINNs are trained in a supervised learning framework to approximate in vitro cell biology assay experiments while respecting …
Article167 citationsarXiv:2005.13073doi:10.1371/journal.pcbi.1008462 - Knowledge-primed neural networks enable biologically interpretable deep learning on single-cell sequencing data
BACKGROUND: Deep learning has emerged as a versatile approach for predicting complex biological phenomena. However, its utility for biological discovery has so far been limited, given that generic deep neural networks provide little insight into the biological mechanisms that underlie a successful prediction. Here we demonstrate deep learning on biological n…
Abstract excerpt by Nikolaus Fortelny et al., CC-BY, https://doi.org/10.1186/s13059-020-02100-5; truncated.
Article135 citationsdoi:10.1186/s13059-020-02100-5 - Machine learning for modeling N2O emissions from wastewater treatment plants: Aligning model performance, complexity, and interpretabilityArticle116 citationsdoi:10.1016/j.watres.2023.120667
- Improving hydrologic models for predictions and process understanding using neural ODEs
Abstract. Deep learning methods have frequently outperformed conceptual hydrologic models in rainfall-runoff modelling. Attempts of investigating the internals of such deep learning models are being made but traceability of model states and processes and their interrelations to model input and output is not yet fully given. Direct interpretability of mechani…
Abstract excerpt by Marvin Höge et al., CC-BY, https://doi.org/10.5194/hess-26-5085-2022; truncated.
Article126 citationsdoi:10.5194/hess-26-5085-2022 - Improving accuracy of rutting prediction for mechanistic-empirical pavement design guide with deep neural networksArticle130 citationsdoi:10.1016/j.conbuildmat.2018.09.087
- Generative Retrieval-Augmented Ontologic Graph and Multiagent Strategies for Interpretive Large Language Model-Based Materials Design
Transformer neural networks show promising capabilities, in particular for uses in materials analysis, design and manufacturing, including their capacity to work effectively with both human language, symbols, code, and numerical data. Here we explore the use of large language models (LLMs) as a tool that can support engineering analysis of materials, applied…
Article82 citationsarXiv:2310.19998doi:10.1021/acsengineeringau.3c00058 - Mechanistic Differences in the Transcriptional Interpretation of Local and Long-Range Shh Morphogen SignalingArticle132 citationsdoi:10.1016/j.devcel.2012.09.015
- Development of a hybrid model for a partially known intracellular signaling pathway through correction term estimation and neural network modeling
Developing an accurate first-principle model is an important step in employing systems biology approaches to analyze an intracellular signaling pathway. However, an accurate first-principle model is difficult to be developed since it requires in-depth mechanistic understandings of the signaling pathway. Since underlying mechanisms such as the reaction networ…
Abstract excerpt by Dongheon Lee et al., CC-BY, https://doi.org/10.1371/journal.pcbi.1008472; truncated.
Article103 citationsdoi:10.1371/journal.pcbi.1008472 - A review of mechanistic learning in mathematical oncology
Mechanistic learning refers to the synergistic combination of mechanistic mathematical modeling and data-driven machine or deep learning. This emerging field finds increasing applications in (mathematical) oncology. This review aims to capture the current state of the field and provides a perspective on how mechanistic learning may progress in the oncology d…
Abstract excerpt by John Metzcar et al., CC-BY, https://doi.org/10.3389/fimmu.2024.1363144; truncated.
Review70 citationsdoi:10.3389/fimmu.2024.1363144 - Efficient inverse graphics in biological face processingArticle89 citationsdoi:10.1126/sciadv.aax5979
- Explainable AI: A review of applications to neuroimaging data
Deep neural networks (DNNs) have transformed the field of computer vision and currently constitute some of the best models for representations learned via hierarchical processing in the human brain. In medical imaging, these models have shown human-level performance and even higher in the early diagnosis of a wide range of diseases. However, the goal is ofte…
Abstract excerpt by Farzad V. Farahani et al., CC-BY, https://doi.org/10.3389/fnins.2022.906290; truncated.
Review81 citationsdoi:10.3389/fnins.2022.906290 - ToxAIcology - The evolving role of artificial intelligence in advancing toxicology and modernizing regulatory science
Toxicology has undergone a transformation from an observational science to a data-rich discipline ripe for artificial intelligence (AI) integration. The exponential growth in computing power coupled with accumulation of large toxicological datasets has created new opportunities to apply techniques like machine learning and especially deep learning to enhance…
Abstract excerpt by Thomas Härtung, CC-BY, https://doi.org/10.14573/altex.2309191; truncated.
Article77 citationsdoi:10.14573/altex.2309191
Newest work
- The Prestige: Benchmarking Cognitive Visual Reasoning using Magic TricksPreprintarXiv:2610.38103
- Replaying the Same Gradients at a Different $\varepsilon$: A Registered Test of the Fixed-Trajectory Account of Adam's $\varepsilon$ for One Rare RowPreprintarXiv:2610.38169
- Scaling Lean Theorem Proving for Real-World Code VerificationPreprintarXiv:2610.38174
- Barren Plateaus in Quantum Linear Combinations: Interference and Lie-Algebraic ControlPreprintarXiv:2610.38193
- From Natural Language to Real Schematics: An Automated Workflow for Power Electronics Schematic SynthesisPreprintarXiv:2610.38201
- Robust LLM-Generated Text Detection through Matched Conditional-Response ProbingPreprintarXiv:2610.38312
- Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability StudyPreprintarXiv:2610.38326
- SpaceAct: Benchmarking 3D Spatial and Temporal Reasoning via Closed-Loop Action and Active PerceptionPreprintarXiv:2610.38405
- What Hallucinations Have We Measured? Semantic Illusion, Confabulation, and Expressed UncertaintyPreprintarXiv:2610.38520
- Value Probes: Learned Value Functions for Beam-Search Red TeamingPreprintarXiv:2610.38526