Translating mass spectra into chemical structures is a central challenge in exposomics, making it difficult to quickly track the millions of chemicals found in humans and the environment. Unlike metabolomics, key problems in developing models for chemicals with a larger molecular space include data scarcity, model complexity and proper query strategy. Here we present a molecular structure generator (MSGo) that can generate structures directly from mass spectra and discover unknown polyfluorinated chemicals in the exposome. Trained with only virtual spectra using a transformer neural network, MSGo correctly identified 48% of structures in a validation set and was better at discovering new polyfluorinated chemicals in wastewater samples reported in the literature than experts. Applying probability-oriented masking to the virtual spectra is key to MSGo’s performance. Rapid discovery of chemicals with limited experimental mass spectral data using automated tools such as MSGo is key to tackling the current unknown polyfluorinated chemical crisis. Yu and colleagues present MSGo, an artificial intelligence exposomics tool trained on virtual mass spectra with masking that identifies pollutants by generating chemical structures that match measured spectral data.
{"title":"Pseudodata-based molecular structure generator to reveal unknown chemicals","authors":"Nanyang Yu, Zheng Ma, Qi Shao, Laihui Li, Xuebing Wang, Bingcai Pan, Hongxia Yu, Si Wei","doi":"10.1038/s42256-025-01140-5","DOIUrl":"10.1038/s42256-025-01140-5","url":null,"abstract":"Translating mass spectra into chemical structures is a central challenge in exposomics, making it difficult to quickly track the millions of chemicals found in humans and the environment. Unlike metabolomics, key problems in developing models for chemicals with a larger molecular space include data scarcity, model complexity and proper query strategy. Here we present a molecular structure generator (MSGo) that can generate structures directly from mass spectra and discover unknown polyfluorinated chemicals in the exposome. Trained with only virtual spectra using a transformer neural network, MSGo correctly identified 48% of structures in a validation set and was better at discovering new polyfluorinated chemicals in wastewater samples reported in the literature than experts. Applying probability-oriented masking to the virtual spectra is key to MSGo’s performance. Rapid discovery of chemicals with limited experimental mass spectral data using automated tools such as MSGo is key to tackling the current unknown polyfluorinated chemical crisis. Yu and colleagues present MSGo, an artificial intelligence exposomics tool trained on virtual mass spectra with masking that identifies pollutants by generating chemical structures that match measured spectral data.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1879-1887"},"PeriodicalIF":23.9,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145509633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-13DOI: 10.1038/s42256-025-01145-0
Kristian Kersting
A new benchmark, KaBLE (knowledge and belief language evaluation), indicates that some large language models are unable to accurately distinguish belief from knowledge and fact, calling into question their use in real-word applications such as medicine and law.
{"title":"Large language models still struggle with false beliefs","authors":"Kristian Kersting","doi":"10.1038/s42256-025-01145-0","DOIUrl":"10.1038/s42256-025-01145-0","url":null,"abstract":"A new benchmark, KaBLE (knowledge and belief language evaluation), indicates that some large language models are unable to accurately distinguish belief from knowledge and fact, calling into question their use in real-word applications such as medicine and law.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1778-1779"},"PeriodicalIF":23.9,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145498438","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-13DOI: 10.1038/s42256-025-01143-2
Bradley H. Theilman, James B. Aimone
The finite element method (FEM) is one of the most important and ubiquitous numerical methods for solving partial differential equations (PDEs) on computers for scientific and engineering discovery. Applying the FEM to larger and more detailed scientific models has driven advances in high-performance computing for decades. Here we demonstrate that scalable spiking neuromorphic hardware can directly implement the FEM by constructing a spiking neural network that solves the large, sparse, linear systems of equations at the core of the FEM. We show that for the Poisson equation, a fundamental PDE in science and engineering, our neural circuit achieves meaningful levels of numerical accuracy and close to ideal scaling on modern, inherently parallel and energy-efficient neuromorphic hardware, specifically Intel’s Loihi 2 neuromorphic platform. We illustrate extensions to irregular mesh geometries in both two and three dimensions as well as other PDEs such as linear elasticity. Our spiking neural network is constructed from a recurrent network model of the brain’s motor cortex and, in contrast to black-box deep artificial neural network-based methods for PDEs, directly translates the well-understood and trusted mathematics of the FEM to a natively spiking neuromorphic algorithm. Theilman and Aimone introduce a natively spiking algorithm for solving partial differential equations on large-scale neuromorphic computers and demonstrate the algorithm on Intel’s Loihi 2 neuromorphic research chip.
{"title":"Solving sparse finite element problems on neuromorphic hardware","authors":"Bradley H. Theilman, James B. Aimone","doi":"10.1038/s42256-025-01143-2","DOIUrl":"10.1038/s42256-025-01143-2","url":null,"abstract":"The finite element method (FEM) is one of the most important and ubiquitous numerical methods for solving partial differential equations (PDEs) on computers for scientific and engineering discovery. Applying the FEM to larger and more detailed scientific models has driven advances in high-performance computing for decades. Here we demonstrate that scalable spiking neuromorphic hardware can directly implement the FEM by constructing a spiking neural network that solves the large, sparse, linear systems of equations at the core of the FEM. We show that for the Poisson equation, a fundamental PDE in science and engineering, our neural circuit achieves meaningful levels of numerical accuracy and close to ideal scaling on modern, inherently parallel and energy-efficient neuromorphic hardware, specifically Intel’s Loihi 2 neuromorphic platform. We illustrate extensions to irregular mesh geometries in both two and three dimensions as well as other PDEs such as linear elasticity. Our spiking neural network is constructed from a recurrent network model of the brain’s motor cortex and, in contrast to black-box deep artificial neural network-based methods for PDEs, directly translates the well-understood and trusted mathematics of the FEM to a natively spiking neuromorphic algorithm. Theilman and Aimone introduce a natively spiking algorithm for solving partial differential equations on large-scale neuromorphic computers and demonstrate the algorithm on Intel’s Loihi 2 neuromorphic research chip.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1845-1857"},"PeriodicalIF":23.9,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.comhttps://www.nature.com/articles/s42256-025-01143-2.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145498437","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-13DOI: 10.1038/s42256-025-01144-1
Mohammad Nadeem, Shahab Saquib Sohail, Erik Cambria, Shagufta Afreen
Biases in artificial intelligence models have been studied predominantly through Western lenses, overlooking South Asia’s unique contexts of caste, religion, colourism and representation. This Comment highlights region-specific biases in language and vision models and calls for fairness frameworks grounded in South Asian realities.
{"title":"South Asian biases in language and vision models","authors":"Mohammad Nadeem, Shahab Saquib Sohail, Erik Cambria, Shagufta Afreen","doi":"10.1038/s42256-025-01144-1","DOIUrl":"10.1038/s42256-025-01144-1","url":null,"abstract":"Biases in artificial intelligence models have been studied predominantly through Western lenses, overlooking South Asia’s unique contexts of caste, religion, colourism and representation. This Comment highlights region-specific biases in language and vision models and calls for fairness frameworks grounded in South Asian realities.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1775-1777"},"PeriodicalIF":23.9,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145498436","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-13DOI: 10.1038/s42256-025-01142-3
Atlas Kazemian, Eric Elmoznino, Michael F. Bonner
What underlies the emergence of cortex-aligned representations in deep neural network models of vision? Earlier work suggested that shared architectural constraints were a major factor, but the success of widely varied architectures after pretraining raises critical questions about the importance of architectural constraints. Here we show that in wide networks with minimal training, architectural inductive biases have a prominent role. We examined networks with varied architectures but no pretraining and quantified their ability to predict image representations in the visual cortices of monkeys and humans. We found that cortex-aligned representations emerge in convolutional architectures that combine two key manipulations of dimensionality: compression in the spatial domain, through pooling, and expansion in the feature domain by increasing the number of channels. We further show that the inductive biases of convolutional architectures are critical for obtaining performance gains from feature expansion—dimensionality manipulations were relatively ineffective in other architectures and in convolutional models with targeted lesions. Our findings suggest that the architectural constraints of convolutional networks are sufficiently close to the constraints of biological vision to allow many aspects of cortical visual representation to emerge even before synaptic connections have been tuned through experience. Kazemian et al. report that untrained convolutional networks with wide layers predict primate visual cortex responses nearly as well as task-optimized networks, revealing how architectural constraints shape brain-like representations in deep networks.
{"title":"Convolutional architectures are cortex-aligned de novo","authors":"Atlas Kazemian, Eric Elmoznino, Michael F. Bonner","doi":"10.1038/s42256-025-01142-3","DOIUrl":"10.1038/s42256-025-01142-3","url":null,"abstract":"What underlies the emergence of cortex-aligned representations in deep neural network models of vision? Earlier work suggested that shared architectural constraints were a major factor, but the success of widely varied architectures after pretraining raises critical questions about the importance of architectural constraints. Here we show that in wide networks with minimal training, architectural inductive biases have a prominent role. We examined networks with varied architectures but no pretraining and quantified their ability to predict image representations in the visual cortices of monkeys and humans. We found that cortex-aligned representations emerge in convolutional architectures that combine two key manipulations of dimensionality: compression in the spatial domain, through pooling, and expansion in the feature domain by increasing the number of channels. We further show that the inductive biases of convolutional architectures are critical for obtaining performance gains from feature expansion—dimensionality manipulations were relatively ineffective in other architectures and in convolutional models with targeted lesions. Our findings suggest that the architectural constraints of convolutional networks are sufficiently close to the constraints of biological vision to allow many aspects of cortical visual representation to emerge even before synaptic connections have been tuned through experience. Kazemian et al. report that untrained convolutional networks with wide layers predict primate visual cortex responses nearly as well as task-optimized networks, revealing how architectural constraints shape brain-like representations in deep networks.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1834-1844"},"PeriodicalIF":23.9,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145498435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-06DOI: 10.1038/s42256-025-01130-7
Yulin Wang (, ), Yang Yue (, ), Yang Yue (, ), Huanqian Wang, Haojun Jiang, Yizeng Han, Zanlin Ni, Yifan Pu, Minglei Shi, Rui Lu, Qisen Yang, Andrew Zhao, Zhuofan Xia, Shiji Song, Gao Huang
Human vision is highly adaptive, efficiently sampling intricate environments by sequentially fixating on task-relevant regions. In contrast, prevailing machine vision models passively process entire scenes at once, resulting in excessive resource demands scaling with spatial–temporal input resolution and model size, yielding critical limitations impeding both future advancements and real-world application. Here we introduce AdaptiveNN, a general framework aiming to enable the transition from ‘passive’ to ‘active and adaptive’ vision models. AdaptiveNN formulates visual perception as a coarse-to-fine sequential decision-making process, progressively identifying and attending to regions pertinent to the task, incrementally combining information across fixations and actively concluding observation when sufficient. We establish a theory integrating representation learning with self-rewarding reinforcement learning, enabling end-to-end training of the non-differentiable AdaptiveNN without additional supervision on fixation locations. We assess AdaptiveNN on 17 benchmarks spanning 9 tasks, including large-scale visual recognition, fine-grained discrimination, visual search, processing images from real driving and medical scenarios, language-driven embodied artificial intelligence and side-by-side comparisons with humans. AdaptiveNN achieves up to 28 times inference cost reduction without sacrificing accuracy, flexibly adapts to varying task demands and resource budgets without retraining, and provides enhanced interpretability via its fixation patterns, demonstrating a promising avenue towards efficient, flexible and interpretable computer vision. Furthermore, AdaptiveNN exhibits closely human-like perceptual behaviours in many cases, revealing its potential as a valuable tool for investigating visual cognition. A deep learning approach, AdaptiveNN, shifts machine vision models from passive to active to mimic human-like perception. The method achieves inference costs that are up to 28-times lower without accuracy loss, while showcasing online-adaptable and interpretable behaviours.
{"title":"Emulating human-like adaptive vision for efficient and flexible machine visual perception","authors":"Yulin Wang \u0000 (, ), Yang Yue \u0000 (, ), Yang Yue \u0000 (, ), Huanqian Wang, Haojun Jiang, Yizeng Han, Zanlin Ni, Yifan Pu, Minglei Shi, Rui Lu, Qisen Yang, Andrew Zhao, Zhuofan Xia, Shiji Song, Gao Huang","doi":"10.1038/s42256-025-01130-7","DOIUrl":"10.1038/s42256-025-01130-7","url":null,"abstract":"Human vision is highly adaptive, efficiently sampling intricate environments by sequentially fixating on task-relevant regions. In contrast, prevailing machine vision models passively process entire scenes at once, resulting in excessive resource demands scaling with spatial–temporal input resolution and model size, yielding critical limitations impeding both future advancements and real-world application. Here we introduce AdaptiveNN, a general framework aiming to enable the transition from ‘passive’ to ‘active and adaptive’ vision models. AdaptiveNN formulates visual perception as a coarse-to-fine sequential decision-making process, progressively identifying and attending to regions pertinent to the task, incrementally combining information across fixations and actively concluding observation when sufficient. We establish a theory integrating representation learning with self-rewarding reinforcement learning, enabling end-to-end training of the non-differentiable AdaptiveNN without additional supervision on fixation locations. We assess AdaptiveNN on 17 benchmarks spanning 9 tasks, including large-scale visual recognition, fine-grained discrimination, visual search, processing images from real driving and medical scenarios, language-driven embodied artificial intelligence and side-by-side comparisons with humans. AdaptiveNN achieves up to 28 times inference cost reduction without sacrificing accuracy, flexibly adapts to varying task demands and resource budgets without retraining, and provides enhanced interpretability via its fixation patterns, demonstrating a promising avenue towards efficient, flexible and interpretable computer vision. Furthermore, AdaptiveNN exhibits closely human-like perceptual behaviours in many cases, revealing its potential as a valuable tool for investigating visual cognition. A deep learning approach, AdaptiveNN, shifts machine vision models from passive to active to mimic human-like perception. The method achieves inference costs that are up to 28-times lower without accuracy loss, while showcasing online-adaptable and interpretable behaviours.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1804-1822"},"PeriodicalIF":23.9,"publicationDate":"2025-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145447396","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-06DOI: 10.1038/s42256-025-01137-0
Chaojun Xiao, Jie Cai, Weilin Zhao, Biyuan Lin, Guoyang Zeng, Jie Zhou, Zhi Zheng, Xu Han, Zhiyuan Liu, Maosong Sun
Large language models (LLMs) have emerged as a milestone in artificial intelligence. The scaling law indicates that the performance of LLMs can continually improve as the model size increases, which poses challenges for training and deployment. Despite numerous efforts to improve LLM efficiency, there is no general consensus on development trends and evaluation metrics for efficiency of LLMs with different scales. To address this tension between model performance and efficiency, we introduce the concept of capability density as a metric to evaluate the quality of the LLMs and describe the trend of LLMs in terms of both effectiveness and efficiency. Intuitively, capability density can be understood as the capability contained within each unit of model parameters. Capability density provides a unified framework for assessing both model performance and efficiency. Here we show an empirical observation, called the ‘densing law’, that the capability density of LLMs grows exponentially over time. More specifically, using widely used benchmarks for evaluation, the maximum capability density of open-source LLMs doubles approximately every 3.5 months. This reveals that both parameter requirements and inference costs of LLMs for achieving equivalent performance decrease exponentially, offering insights for efficient LLM development strategies. Xiao et al. introduce ‘capability density’, defined as capability per parameter, as a metric for evaluating large language models. They report an empirical trend, the ‘densing law’, which states that capability density doubles approximately every 3.5 months, indicating that equivalent model performance can be achieved with exponentially fewer parameters over time.
{"title":"Densing law of LLMs","authors":"Chaojun Xiao, Jie Cai, Weilin Zhao, Biyuan Lin, Guoyang Zeng, Jie Zhou, Zhi Zheng, Xu Han, Zhiyuan Liu, Maosong Sun","doi":"10.1038/s42256-025-01137-0","DOIUrl":"10.1038/s42256-025-01137-0","url":null,"abstract":"Large language models (LLMs) have emerged as a milestone in artificial intelligence. The scaling law indicates that the performance of LLMs can continually improve as the model size increases, which poses challenges for training and deployment. Despite numerous efforts to improve LLM efficiency, there is no general consensus on development trends and evaluation metrics for efficiency of LLMs with different scales. To address this tension between model performance and efficiency, we introduce the concept of capability density as a metric to evaluate the quality of the LLMs and describe the trend of LLMs in terms of both effectiveness and efficiency. Intuitively, capability density can be understood as the capability contained within each unit of model parameters. Capability density provides a unified framework for assessing both model performance and efficiency. Here we show an empirical observation, called the ‘densing law’, that the capability density of LLMs grows exponentially over time. More specifically, using widely used benchmarks for evaluation, the maximum capability density of open-source LLMs doubles approximately every 3.5 months. This reveals that both parameter requirements and inference costs of LLMs for achieving equivalent performance decrease exponentially, offering insights for efficient LLM development strategies. Xiao et al. introduce ‘capability density’, defined as capability per parameter, as a metric for evaluating large language models. They report an empirical trend, the ‘densing law’, which states that capability density doubles approximately every 3.5 months, indicating that equivalent model performance can be achieved with exponentially fewer parameters over time.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1823-1833"},"PeriodicalIF":23.9,"publicationDate":"2025-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.nature.comhttps://www.nature.com/articles/s42256-025-01137-0.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145447372","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-03DOI: 10.1038/s42256-025-01113-8
Mirac Suzgun, Tayfun Gur, Federico Bianchi, Daniel E. Ho, Thomas Icard, Dan Jurafsky, James Zou
As language models (LMs) increasingly infiltrate into high-stakes domains such as law, medicine, journalism and science, their ability to distinguish belief from knowledge, and fact from fiction, becomes imperative. Failure to make such distinctions can mislead diagnoses, distort judicial judgments and amplify misinformation. Here we evaluate 24 cutting-edge LMs using a new KaBLE benchmark of 13,000 questions across 13 epistemic tasks. Our findings reveal crucial limitations. In particular, all models tested systematically fail to acknowledge first-person false beliefs, with GPT-4o dropping from 98.2% to 64.4% accuracy and DeepSeek R1 plummeting from over 90% to 14.4%. Further, models process third-person false beliefs with substantially higher accuracy (95% for newer models; 79% for older ones) than first-person false beliefs (62.6% for newer; 52.5% for older), revealing a troubling attribution bias. We also find that, while recent models show competence in recursive knowledge tasks, they still rely on inconsistent reasoning strategies, suggesting superficial pattern matching rather than robust epistemic understanding. Most models lack a robust understanding of the factive nature of knowledge, that knowledge inherently requires truth. These limitations necessitate urgent improvements before deploying LMs in high-stakes domains where epistemic distinctions are crucial. Suzgun et al. find that current large language models cannot reliably distinguish between belief, knowledge and fact, raising concerns for their use in healthcare, law and journalism, where such distinctions are critical.
{"title":"Language models cannot reliably distinguish belief from knowledge and fact","authors":"Mirac Suzgun, Tayfun Gur, Federico Bianchi, Daniel E. Ho, Thomas Icard, Dan Jurafsky, James Zou","doi":"10.1038/s42256-025-01113-8","DOIUrl":"10.1038/s42256-025-01113-8","url":null,"abstract":"As language models (LMs) increasingly infiltrate into high-stakes domains such as law, medicine, journalism and science, their ability to distinguish belief from knowledge, and fact from fiction, becomes imperative. Failure to make such distinctions can mislead diagnoses, distort judicial judgments and amplify misinformation. Here we evaluate 24 cutting-edge LMs using a new KaBLE benchmark of 13,000 questions across 13 epistemic tasks. Our findings reveal crucial limitations. In particular, all models tested systematically fail to acknowledge first-person false beliefs, with GPT-4o dropping from 98.2% to 64.4% accuracy and DeepSeek R1 plummeting from over 90% to 14.4%. Further, models process third-person false beliefs with substantially higher accuracy (95% for newer models; 79% for older ones) than first-person false beliefs (62.6% for newer; 52.5% for older), revealing a troubling attribution bias. We also find that, while recent models show competence in recursive knowledge tasks, they still rely on inconsistent reasoning strategies, suggesting superficial pattern matching rather than robust epistemic understanding. Most models lack a robust understanding of the factive nature of knowledge, that knowledge inherently requires truth. These limitations necessitate urgent improvements before deploying LMs in high-stakes domains where epistemic distinctions are crucial. Suzgun et al. find that current large language models cannot reliably distinguish between belief, knowledge and fact, raising concerns for their use in healthcare, law and journalism, where such distinctions are critical.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1780-1790"},"PeriodicalIF":23.9,"publicationDate":"2025-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145434427","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-24DOI: 10.1038/s42256-025-01129-0
Ahmed Y. Ismail, Bradley A. A. Martin, Keith T. Butler
Molecular dynamics (MD) simulations are widely used for understanding atomic motion but require substantial computational time. In new research by Nam et al., a generative artificial intelligence framework is developed to accelerate the MD simulations for crystalline materials, by reframing the task as conditional generation of atomic displacement.
{"title":"Accelerating molecular dynamics by going with the flow","authors":"Ahmed Y. Ismail, Bradley A. A. Martin, Keith T. Butler","doi":"10.1038/s42256-025-01129-0","DOIUrl":"10.1038/s42256-025-01129-0","url":null,"abstract":"Molecular dynamics (MD) simulations are widely used for understanding atomic motion but require substantial computational time. In new research by Nam et al., a generative artificial intelligence framework is developed to accelerate the MD simulations for crystalline materials, by reframing the task as conditional generation of atomic displacement.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 10","pages":"1598-1599"},"PeriodicalIF":23.9,"publicationDate":"2025-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145352987","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-23DOI: 10.1038/s42256-025-01136-1
Huanyu Tao, Xiaoyu Wang, Sheng-You Huang
Accurate prediction of protein–peptide interactions is critical for peptide drug discovery. However, due to the limited number of protein–peptide structures in the Protein Data Bank, it is challenging to train an accurate scoring function for protein–peptide interactions. Here, addressing this challenge, we propose an interaction-derived graph neural network model for scoring protein–peptide complexes, named GraphPep. GraphPep models protein–peptide interactions instead of traditional atoms or residues as graph nodes, and focuses on residue–residue contacts instead of a single peptide root mean square deviation in the loss function. Therefore, GraphPep can not only efficiently capture the most important protein–peptide interactions, but also mitigate the problem of limited training data. Moreover, the power of GraphPep is further enhanced by the ESM-2 protein language model. GraphPep is extensively evaluated on diverse decoy sets generated by various protein–peptide docking programs and AlphaFold, and is compared against state-of-the-art methods. The results demonstrate the accuracy and robustness of GraphPep. GraphPep presents an interaction-derived and protein language model-powered graph learning framework for robust scoring of protein–peptide complexes, substantially enhancing the binding mode prediction of protein–peptide docking.
{"title":"An interaction-derived graph learning framework for scoring protein–peptide complexes","authors":"Huanyu Tao, Xiaoyu Wang, Sheng-You Huang","doi":"10.1038/s42256-025-01136-1","DOIUrl":"10.1038/s42256-025-01136-1","url":null,"abstract":"Accurate prediction of protein–peptide interactions is critical for peptide drug discovery. However, due to the limited number of protein–peptide structures in the Protein Data Bank, it is challenging to train an accurate scoring function for protein–peptide interactions. Here, addressing this challenge, we propose an interaction-derived graph neural network model for scoring protein–peptide complexes, named GraphPep. GraphPep models protein–peptide interactions instead of traditional atoms or residues as graph nodes, and focuses on residue–residue contacts instead of a single peptide root mean square deviation in the loss function. Therefore, GraphPep can not only efficiently capture the most important protein–peptide interactions, but also mitigate the problem of limited training data. Moreover, the power of GraphPep is further enhanced by the ESM-2 protein language model. GraphPep is extensively evaluated on diverse decoy sets generated by various protein–peptide docking programs and AlphaFold, and is compared against state-of-the-art methods. The results demonstrate the accuracy and robustness of GraphPep. GraphPep presents an interaction-derived and protein language model-powered graph learning framework for robust scoring of protein–peptide complexes, substantially enhancing the binding mode prediction of protein–peptide docking.","PeriodicalId":48533,"journal":{"name":"Nature Machine Intelligence","volume":"7 11","pages":"1858-1869"},"PeriodicalIF":23.9,"publicationDate":"2025-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145381720","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：