Yarin Gal
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yarin Gal has a total of 287 co-authors. Here are the top co-authors by number of shared publications:
- Debora Susan Marks
Department of Systems Biology, Harvard Medical School, Boston, USA
9 shared publications - Pascal Notin
Department of Systems Biology, Harvard Medical School, Boston, MA
7 shared publications - Sören Mindermann
EECS UC Berkeley University of Oxford
5 shared publications - Andrés Muñoz-Jaramillo
Southwest Research Institute, Boulder, CO 80302, USA
4 shared publications - Atılım Güneş Baydin
Department of Computer Science University of Oxford
4 shared publications - Atılım Güneş Baydin
University of Surrey University of Oxford European Space Agency
4 shared publications
Latest Publications
- 5 October 2026
- Published by Springer Nature
- p. 232-249
FindingDory: A Benchmark to Evaluate Memory in Embodied Agents
- 5 October 2026
- Published by Springer Nature
- p. 232-249
FindingDory: A Benchmark to Evaluate Memory in Embodied Agents
- Open Access
- 11 April 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Hybrid physics-machine learning models for quantitative electron diffraction refinements
High accuracy electron microscopy simulations required for quantitative crystal structure refinements face a fundamental challenge: while physical interactions are well-described theoretically, real-world experimental effects are challenging to model analytically. To address this gap, we present a hybrid physics-machine learning framework that integrates differentiable physical simulations with neural networks. By leveraging automatic differentiation throughout the simulation pipeline, our method enables gradient-based joint optimization of physical parameters and neural network components representing experimental variables, offering superior scalability compared to traditional second-order methods. We demonstrate this framework through application to three-dimensional electron diffraction (3D-ED) structure refinement, where our approach learns complex thickness distributions directly from diffraction data rather than relying on simplified geometric models. This method achieves state-of-the-art refinement performance across synthetic and experimental datasets, recovering atomic positions, thermal displacements, and thickness profiles with high fidelity. The modular architecture proposed can naturally be extended to accommodate additional physical phenomena and extended to other electron microscopy techniques. This establishes differentiable hybrid modeling as a powerful paradigm for quantitative electron microscopy, where experimental complexities have historically limited analysis.
- 11 April 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessHybrid physics-machine learning models for quantitative electron diffraction refinements
High accuracy electron microscopy simulations required for quantitative crystal structure refinements face a fundamental challenge: while physical interactions are well-described theoretically, real-world experimental effects are challenging to model analytically. To address this gap, we present a hybrid physics-machine learning framework that integrates differentiable physical simulations with neural networks. By leveraging automatic differentiation throughout the simulation pipeline, our method enables gradient-based joint optimization of physical parameters and neural network components representing experimental variables, offering superior scalability compared to traditional second-order methods. We demonstrate this framework through application to three-dimensional electron diffraction (3D-ED) structure refinement, where our approach learns complex thickness distributions directly from diffraction data rather than relying on simplified geometric models. This method achieves state-of-the-art refinement performance across synthetic and experimental datasets, recovering atomic positions, thermal displacements, and thickness profiles with high fidelity. The modular architecture proposed can naturally be extended to accommodate additional physical phenomena and extended to other electron microscopy techniques. This establishes differentiable hybrid modeling as a powerful paradigm for quantitative electron microscopy, where experimental complexities have historically limited analysis.
- 7 October 2025
- Published by Springer Nature in Nature
- Vol. 646 (8084) , 286-287
Customizable AI systems that anyone can adapt bring big opportunities — and even bigger risks
- 7 October 2025
- Published by Springer Nature in Nature
- Vol. 646 (8084) , 286-287
Customizable AI systems that anyone can adapt bring big opportunities — and even bigger risks
- Open Access
Sampling Protein Language Models for Functional Protein Design
Protein language models have emerged as powerful tools for learning rich protein representations, improving performance in tasks like structure prediction, mutation effect estimation, and homology detection. Their ability to model complex sequence distributions also holds promise for designing novel, functional proteins with broad applications in therapeutics, materials, and sustainability. However, due to the vastness of sequence space, efficient exploration methods are essential for protein engineering. Despite this, most existing protein design approaches using protein language models rely on single-mutant sampling strategies borrowed from Natural Language Processing, which fail to capture critical epistatic interactions between amino acid positions that are essential for protein function. In this work, we develop a comprehensive in silico protein design evaluation framework to systematically compare different sampling methods. After a thorough review of existing sampling strategies for language models, we introduce several approaches specifically tailored for protein design. We demonstrate that sampling strategies that consider multiple mutations simultaneously significantly outperform single-mutant approaches by better capturing epistatic effects between residue pairs. We evaluated these strategies using our framework, investigating the effects of key hyperparameters and providing practical guidance on the relative strengths of each method depending on design objectives.
Open AccessSampling Protein Language Models for Functional Protein Design
Protein language models have emerged as powerful tools for learning rich protein representations, improving performance in tasks like structure prediction, mutation effect estimation, and homology detection. Their ability to model complex sequence distributions also holds promise for designing novel, functional proteins with broad applications in therapeutics, materials, and sustainability. However, due to the vastness of sequence space, efficient exploration methods are essential for protein engineering. Despite this, most existing protein design approaches using protein language models rely on single-mutant sampling strategies borrowed from Natural Language Processing, which fail to capture critical epistatic interactions between amino acid positions that are essential for protein function. In this work, we develop a comprehensive in silico protein design evaluation framework to systematically compare different sampling methods. After a thorough review of existing sampling strategies for language models, we introduce several approaches specifically tailored for protein design. We demonstrate that sampling strategies that consider multiple mutations simultaneously significantly outperform single-mutant approaches by better capturing epistatic effects between residue pairs. We evaluated these strategies using our framework, investigating the effects of key hyperparameters and providing practical guidance on the relative strengths of each method depending on design objectives.
- Open Access
Hybrid Physics-Machine Learning Models for Quantitative Electron Diffraction Refinements
High-fidelity electron microscopy simulations required for quantitative crystal structure refinements face a fundamental challenge: while physical interactions are well-described theoretically, real-world experimental effects are challenging to model analytically. To address this gap, we present a novel hybrid physics-machine learning framework that integrates differentiable physical simulations with neural networks. By leveraging automatic differentiation throughout the simulation pipeline, our method enables gradient-based joint optimization of physical parameters and neural network components representing experimental variables, offering superior scalability compared to traditional second-order methods. We demonstrate this framework through application to three-dimensional electron diffraction (3D-ED) structure refinement, where our approach learns complex thickness distributions directly from diffraction data rather than relying on simplified geometric models. This method achieves state-of-the-art refinement performance across synthetic and experimental datasets, recovering atomic positions, thermal displacements, and thickness profiles with high fidelity. The modular architecture proposed can naturally be extended to accommodate additional physical phenomena and extended to other electron microscopy techniques. This establishes differentiable hybrid modeling as a powerful new paradigm for quantitative electron microscopy, where experimental complexities have historically limited analysis.
Open AccessHybrid Physics-Machine Learning Models for Quantitative Electron Diffraction Refinements
High-fidelity electron microscopy simulations required for quantitative crystal structure refinements face a fundamental challenge: while physical interactions are well-described theoretically, real-world experimental effects are challenging to model analytically. To address this gap, we present a novel hybrid physics-machine learning framework that integrates differentiable physical simulations with neural networks. By leveraging automatic differentiation throughout the simulation pipeline, our method enables gradient-based joint optimization of physical parameters and neural network components representing experimental variables, offering superior scalability compared to traditional second-order methods. We demonstrate this framework through application to three-dimensional electron diffraction (3D-ED) structure refinement, where our approach learns complex thickness distributions directly from diffraction data rather than relying on simplified geometric models. This method achieves state-of-the-art refinement performance across synthetic and experimental datasets, recovering atomic positions, thermal displacements, and thickness profiles with high fidelity. The modular architecture proposed can naturally be extended to accommodate additional physical phenomena and extended to other electron microscopy techniques. This establishes differentiable hybrid modeling as a powerful new paradigm for quantitative electron microscopy, where experimental complexities have historically limited analysis.
- Open Access
- 19 June 2025
- Published by American Astronomical Society in The Astronomical Journal
- Vol. 170 (1) , 39
Accelerating Long-period Exoplanet Discovery by Combining Deep Learning and Citizen Science
Automated planetary transit detection has become vital to identify and prioritize candidates for expert analysis and verification given the scale of modern telescopic surveys. Current methods for short-period exoplanet detection work effectively due to periodicity in the transit signals, but a robust approach for detecting single-transit events is lacking. However, volunteer-labeled transits collected by the Planet Hunters TESS (PHT) project now provide an unprecedented opportunity to investigate a data-driven approach to long-period exoplanet detection. In this work, we train a 1D convolutional neural network to classify planetary transits using PHT volunteer scores as training data. We find that this model recovers planet candidates (TESS objects of interest; TOIs) at a precision and recall rate exceeding those of volunteers, with a 20% improvement in the area under the precision-recall curve and 10% more TOIs identified in the top 500 predictions on average per sector. Importantly, the model also recovers almost all planet candidates found by volunteers but missed by current automated methods (PHT community TOIs). Finally we retrospectively utilise the model to simulate live deployment in PHT to reprioritize candidates for analysis. We also find that multiple promising planet candidates, originally missed by PHT, would have been found using our approach, showing promise for upcoming real-world deployment.
- 19 June 2025
- Published by American Astronomical Society in The Astronomical Journal
- Vol. 170 (1) , 39
Open AccessAccelerating Long-period Exoplanet Discovery by Combining Deep Learning and Citizen Science
Automated planetary transit detection has become vital to identify and prioritize candidates for expert analysis and verification given the scale of modern telescopic surveys. Current methods for short-period exoplanet detection work effectively due to periodicity in the transit signals, but a robust approach for detecting single-transit events is lacking. However, volunteer-labeled transits collected by the Planet Hunters TESS (PHT) project now provide an unprecedented opportunity to investigate a data-driven approach to long-period exoplanet detection. In this work, we train a 1D convolutional neural network to classify planetary transits using PHT volunteer scores as training data. We find that this model recovers planet candidates (TESS objects of interest; TOIs) at a precision and recall rate exceeding those of volunteers, with a 20% improvement in the area under the precision-recall curve and 10% more TOIs identified in the top 500 predictions on average per sector. Importantly, the model also recovers almost all planet candidates found by volunteers but missed by current automated methods (PHT community TOIs). Finally we retrospectively utilise the model to simulate live deployment in PHT to reprioritize candidates for analysis. We also find that multiple promising planet candidates, originally missed by PHT, would have been found using our approach, showing promise for upcoming real-world deployment.
- Open Access
Protriever: End-to-End Differentiable Protein Homology Search for Fitness Prediction
Retrieving homologous protein sequences is essential for a broad range of protein modeling tasks such as fitness prediction, protein design, structure modeling, and protein-protein interactions. Traditional workflows have relied on a two-step process: first retrieving homologs via Multiple Sequence Alignments (MSA), then training models on one or more of these alignments. However, MSA-based retrieval is computationally expensive, struggles with highly divergent sequences or complex insertions & deletions patterns, and operates independently of the downstream modeling objective. We introduce Protriever, an end-to-end differentiable framework that learns to retrieve relevant homologs while simultaneously training for the target task. When applied to protein fitness prediction, Protriever achieves state-of-the-art performance compared to sequence-based models that rely on MSA-based homolog retrieval, while being two orders of magnitude faster through efficient vector search. Protriever is both architecture- and task-agnostic, and can flexibly adapt to different retrieval strategies and protein databases at inference time -- offering a scalable alternative to alignment-centric approaches.
Open AccessProtriever: End-to-End Differentiable Protein Homology Search for Fitness Prediction
Retrieving homologous protein sequences is essential for a broad range of protein modeling tasks such as fitness prediction, protein design, structure modeling, and protein-protein interactions. Traditional workflows have relied on a two-step process: first retrieving homologs via Multiple Sequence Alignments (MSA), then training models on one or more of these alignments. However, MSA-based retrieval is computationally expensive, struggles with highly divergent sequences or complex insertions & deletions patterns, and operates independently of the downstream modeling objective. We introduce Protriever, an end-to-end differentiable framework that learns to retrieve relevant homologs while simultaneously training for the target task. When applied to protein fitness prediction, Protriever achieves state-of-the-art performance compared to sequence-based models that rely on MSA-based homolog retrieval, while being two orders of magnitude faster through efficient vector search. Protriever is both architecture- and task-agnostic, and can flexibly adapt to different retrieval strategies and protein databases at inference time -- offering a scalable alternative to alignment-centric approaches.
- Open Access
- 21 March 2025
- Published by Springer Nature in Nature
- Vol. 640 (8058) , E6
Author Correction: AI models collapse when trained on recursively generated data
- 21 March 2025
- Published by Springer Nature in Nature
- Vol. 640 (8058) , E6
Open AccessAuthor Correction: AI models collapse when trained on recursively generated data
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 32298-32310
Detecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions
Hazel Kim, Tom A. Lamb, Adel Bibi, Philip Torr, Yarin Gal. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 32298-32310
Open AccessDetecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions
Hazel Kim, Tom A. Lamb, Adel Bibi, Philip Torr, Yarin Gal. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- 19 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-11
Generating Physically-Consistent Satellite Imagery for Climate Visualizations
Deep generative vision models are now able to synthesize realistic-looking satellite imagery. But, the possibility of hallucinations prevents their adoption for risk-sensitive applications, such as generating materials for communicating climate change. To demonstrate this issue, we train a generative adversarial network (pix2pixHD) to create synthetic satellite imagery of future flooding and reforestation events. We find that a pure deep learning-based model can generate photorealistic flood visualizations but hallucinates floods at locations that were not susceptible to flooding. To address this issue, we propose to condition and evaluate generative vision models on segmentation maps of physics-based flood models. We show that our physics-conditioned model outperforms the pure deep learning-based model and a handcrafted baseline. We evaluate the generalization capability of our method to different remote sensing data and different climate-related events (reforestation). We publish our code and dataset which includes the data for a third case study of melting Arctic sea ice and >30,000 labeled HD image triplets - or the equivalent of 5.5 million images at 128x128 pixels - for segmentation guided image-to-image translation in Earth observation. Code and data is available at github.com/blutjens/eieearth-public.
- 19 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-11
Generating Physically-Consistent Satellite Imagery for Climate Visualizations
Deep generative vision models are now able to synthesize realistic-looking satellite imagery. But, the possibility of hallucinations prevents their adoption for risk-sensitive applications, such as generating materials for communicating climate change. To demonstrate this issue, we train a generative adversarial network (pix2pixHD) to create synthetic satellite imagery of future flooding and reforestation events. We find that a pure deep learning-based model can generate photorealistic flood visualizations but hallucinates floods at locations that were not susceptible to flooding. To address this issue, we propose to condition and evaluate generative vision models on segmentation maps of physics-based flood models. We show that our physics-conditioned model outperforms the pure deep learning-based model and a handcrafted baseline. We evaluate the generalization capability of our method to different remote sensing data and different climate-related events (reforestation). We publish our code and dataset which includes the data for a third case study of melting Arctic sea ice and >30,000 labeled HD image triplets - or the equivalent of 5.5 million images at 128x128 pixels - for segmentation guided image-to-image translation in Earth observation. Code and data is available at github.com/blutjens/eieearth-public.