{ "generated_at": "2026-04-09T14:51:56.329269+08:00", "timezone": "Asia/Shanghai", "lookback_hours": 24, "highlights": [ "主题「Language Model」：命中 4 篇，覆盖 PubMed AI，代表论文包括 《Subcategory vs category fluency: Items and networks in healthy young adults and simulation with a large language model.》、《Sequence Display enables large-scale sequence-activity datasets for rapid protein evolution.》。", "主题「Benchmark」：命中 2 篇，覆盖 PubMed AI，代表论文包括 《ClinicRealm: Re-evaluating large language models with conventional machine learning for non-generative clinical prediction tasks.》、《A guide to using embedded ethics in human stem-cell-based embryo model research.》。", "主题「Evaluation」：命中 1 篇，覆盖 PubMed AI，代表论文包括 《A guide to using embedded ethics in human stem-cell-based embryo model research.》。" ], "topic_sections": [ { "name": "Language Model", "paper_count": 4, "feed_names": [ "PubMed AI" ], "paper_titles": [ "Subcategory vs category fluency: Items and networks in healthy young adults and simulation with a large language model.", "Sequence Display enables large-scale sequence-activity datasets for rapid protein evolution.", "ClinicRealm: Re-evaluating large language models with conventional machine learning for non-generative clinical prediction tasks.", "Advancing neurotech justice in youth digital mental health: insights from an interdisciplinary and cross-generational workshop." ], "key_points": [ "《Subcategory vs category fluency: Items and networks in healthy young adults and simulation with a large language model.》〔评测 / 应用 / 方法〕：Category fluency tasks involve producing words constrained by a semantic field (animals). Subcategory fluency involves producing words from categories that are…", "《Sequence Display enables large-scale sequence-activity datasets for rapid protein evolution.》〔数据 / 应用 / 方法〕：Engineering proteins with desired functions remains challenging and usually requires multiple rounds of screening and selection. Here, we present Sequence Disp…" ] }, { "name": "Benchmark", "paper_count": 2, "feed_names": [ "PubMed AI" ], "paper_titles": [ "ClinicRealm: Re-evaluating large language models with conventional machine learning for non-generative clinical prediction tasks.", "A guide to using embedded ethics in human stem-cell-based embryo model research." ], "key_points": [ "《ClinicRealm: Re-evaluating large language models with conventional machine learning for non-generative clinical prediction tasks.》〔评测 / 方法〕：Large Language Models (LLMs) are increasingly deployed in medicine. However, their utility for non-generative clinical prediction is under-evaluated, and they…", "《A guide to using embedded ethics in human stem-cell-based embryo model research.》〔评测 / 应用 / 方法〕：Human stem-cell-based embryo models (hSCBEMs) offer unprecedented opportunities for basic and translational research. However, the rapid pace of scientific dev…" ] }, { "name": "Evaluation", "paper_count": 1, "feed_names": [ "PubMed AI" ], "paper_titles": [ "A guide to using embedded ethics in human stem-cell-based embryo model research." ], "key_points": [ "《A guide to using embedded ethics in human stem-cell-based embryo model research.》〔评测 / 应用 / 方法〕：Human stem-cell-based embryo models (hSCBEMs) offer unprecedented opportunities for basic and translational research. However, the rapid pace of scientific dev…" ] } ], "template": "zh_daily_brief", "feeds": [ { "name": "LLM", "key_points": [], "papers": [] }, { "name": "Vision", "key_points": [], "papers": [] }, { "name": "PubMed AI", "key_points": [ "《Subcategory vs category fluency: Items and networks in healthy young adults and simulation with a large language model.》〔评测 / 应用 / 方法〕：Category fluency tasks involve producing words constrained by a semantic field (animals). Subcategory fluency involves producing words from categories that are…", "《Sequence Display enables large-scale sequence-activity datasets for rapid protein evolution.》〔数据 / 应用 / 方法〕：Engineering proteins with desired functions remains challenging and usually requires multiple rounds of screening and selection. Here, we present Sequence Disp…", "《ClinicRealm: Re-evaluating large language models with conventional machine learning for non-generative clinical prediction tasks.》〔评测 / 方法〕：Large Language Models (LLMs) are increasingly deployed in medicine. However, their utility for non-generative clinical prediction is under-evaluated, and they…" ], "papers": [ { "title": "Subcategory vs category fluency: Items and networks in healthy young adults and simulation with a large language model.", "summary": "Category fluency tasks involve producing words constrained by a semantic field (animals). Subcategory fluency involves producing words from categories that are semantically related to a superordinate category but form a restricted set of items (farm animals). Here, we study whether people produce different patterns of words in category versus subcategory fluency by looking at differences in the total number of words produced, the properties of the words produced (e.g., frequency) and how people group words together (clusters/switches and network metrics). Forty-eight Dutch-speaking university students responded to three category fluency tasks (animals, foods, transport) and three subcategory fluency tasks (farm animals, fruits, bike parts). Also, we queried a large language model (LLM) to provide responses for 50 \"pseudo-participants\" for the same six categories. People in category (versus subcategory) tasks produced more words; words of higher frequency, with fewer orthographic and phonological neighbors, and shorter in length. They also produced fewer cluster switches and bigger clusters. The category and subcategory networks had different structure (e.g., number of nodes, edges, clustering coefficient). With the LLM we simulated the results regarding word properties and cluster size, but found differences regarding correct words, number of switches, and overlapping clusters between foods and fruit fluency. The differences between category and subcategory fluency may stem from differences in mental search in the lexico-semantic system. However, category and subcategory fluency tasks may be different tasks altogether. The LLM simulation provides novel insights (e.g., how words relate, task-order effects) and suggests caution when used to understand human fluency data.", "authors": [ "Adrià Rofes", "Demi van Dijk", "Jeffrey C Zemla" ], "categories": [ "Journal Article" ], "paper_id": "pubmed:41951933", "abstract_url": "https://pubmed.ncbi.nlm.nih.gov/41951933/", "pdf_url": null, "published_at": "2026-04-08T23:28:00+00:00", "updated_at": "2026-04-08T23:28:00+00:00", "source": "pubmed", "date_label": "Entered", "analysis": null, "tags": [ "评测", "应用", "方法" ], "topics": [ "Language Model" ] }, { "title": "Sequence Display enables large-scale sequence-activity datasets for rapid protein evolution.", "summary": "Engineering proteins with desired functions remains challenging and usually requires multiple rounds of screening and selection. Here, we present Sequence Display, a platform that generates large-scale protein sequence-activity datasets in a single round. Sequence Display enables multiplexed assessment of individual variant activity within a single experiment, offering a robust approach to mapping detailed sequence-function relationships. We demonstrate the platform's broad applicability by generating datasets for cytosine deaminase, uracil glycosylase inhibitor, aminoacyl-tRNA synthetase and a compact Cas9 nuclease. Integrating these datasets obtained from Sequence Display with pretrained protein language models, fine-grained, variant-specific activity landscapes can be constructed. We discovered several Cas9 variants with expanded protospacer-adjacent motif recognition and evolved aminoacyl-tRNA synthetase variants capable of recognizing different noncanonical amino acids. Together, this study establishes Sequence Display as a powerful tool for mapping protein activity landscapes and accelerating the discovery of optimized proteins for biological and medical applications.", "authors": [ "Linqi Cheng", "Xinzhe Zheng", "Shiyu Jason Jiang", "Yu Hu", "Yijie Liu", "Kaiqiang Yang", "Jinyan Rui", "Haoxue Ding", "Mengxi Zhang", "Teng Yuan", "Qianglan Lu", "Haoxin Ye", "Chen-Long Li", "Yiming Guo", "Zuotong Tian", "Anna Qin", "Boyang Zhou", "Kevin K Yang", "Xiongyi Huang", "Han Xiao" ], "categories": [ "Journal Article" ], "paper_id": "pubmed:41951911", "abstract_url": "https://pubmed.ncbi.nlm.nih.gov/41951911/", "pdf_url": null, "published_at": "2026-04-08T23:28:00+00:00", "updated_at": "2026-04-08T23:28:00+00:00", "source": "pubmed", "date_label": "Entered", "analysis": null, "tags": [ "数据", "应用", "方法" ], "topics": [ "Language Model" ] }, { "title": "ClinicRealm: Re-evaluating large language models with conventional machine learning for non-generative clinical prediction tasks.", "summary": "Large Language Models (LLMs) are increasingly deployed in medicine. However, their utility for non-generative clinical prediction is under-evaluated, and they are often assumed to be inferior to specialized models, creating potential for misuse and misunderstanding. To address this, our ClinicRealm benchmark systematically evaluates 15 GPT-style LLMs, 5 BERT-style models, and 11 traditional methods on unstructured clinical notes and structured Electronic Health Records (EHR) across predictive performance, reasoning, fairness, etc. Our findings reveal a significant shift: on clinical notes, leading zero-shot LLMs (e.g., DeepSeek-V3.1-Think, GPT-5) now decisively outperform finetuned BERT models. On structured EHRs, while specialized models excel with ample data, advanced LLMs demonstrate potent zero-shot capabilities, often surpassing conventional models in data-scarce settings. Notably, leading open-source LLMs match or exceed their proprietary counterparts. This provides compelling evidence that modern LLMs are competitive tools for clinical prediction, necessitating a re-evaluation of model selection strategies by health data scientists and developers.", "authors": [ "Yinghao Zhu", "Junyi Gao", "Zixiang Wang", "Weibin Liao", "Xiaochen Zheng", "Lifang Liang", "Miguel O Bernabeu", "Yasha Wang", "Lequan Yu", "Chengwei Pan", "Ewen M Harrison", "Liantao Ma" ], "categories": [ "Journal Article" ], "paper_id": "pubmed:41951858", "abstract_url": "https://pubmed.ncbi.nlm.nih.gov/41951858/", "pdf_url": null, "published_at": "2026-04-08T23:25:00+00:00", "updated_at": "2026-04-08T23:25:00+00:00", "source": "pubmed", "date_label": "Entered", "analysis": null, "tags": [ "评测", "方法" ], "topics": [ "Language Model", "Benchmark" ] }, { "title": "Advancing neurotech justice in youth digital mental health: insights from an interdisciplinary and cross-generational workshop.", "summary": "Researchers and clinicians are increasingly looking to leverage artificial intelligence (AI) and digital tools to improve psychiatric care. Of particular promise is addressing the youth mental health crisis. Yet, the introduction of AI-enabled digital technologies for psychiatric treatment of young adults raises a host of ethical, legal, and societal issues (ELSI). To provide guidance in addressing these issues, we convened a two-day meeting at the Radcliffe Institute for Advanced Study at Harvard University: Advancing Neurotech Justice in Mental Health: Insights from an Interdisciplinary and Cross-Generational Workshop. The meeting brought together a diverse cohort of 17 experts and 5 students from various fields and different countries. In partnership with the MIT Critical Data team, the workshop engaged participants in an interactive Prompt-a-Thon to explore first-hand the potential benefits, biases, and harms related to the use of Large Language Model chatbots in mental health care. This Perspective reports on five principles of digital psychiatry deployment that the workshop participants determined to be the most essential: ensuring accuracy, remaining human-centric, promoting just access, protecting privacy, and providing transparency. We place these five principles within a \"Neurotech Justice\" framework and discuss how guardrails can be built to promote neurotech justice in digital psychiatry.", "authors": [ "Craig W McFarland", "Donnella S Comeau", "Sepideh Abdi", "Mahsa Alborzi Avanaki", "Leo Anthony Celi", "Neurotech Justice Workshop Participants", "Francis X Shen", "Benjamin C Silverman" ], "categories": [ "Journal Article", "Review" ], "paper_id": "pubmed:41951757", "abstract_url": "https://pubmed.ncbi.nlm.nih.gov/41951757/", "pdf_url": null, "published_at": "2026-04-08T23:21:00+00:00", "updated_at": "2026-04-08T23:21:00+00:00", "source": "pubmed", "date_label": "Entered", "analysis": null, "tags": [ "应用", "方法" ], "topics": [ "Language Model" ] }, { "title": "A guide to using embedded ethics in human stem-cell-based embryo model research.", "summary": "Human stem-cell-based embryo models (hSCBEMs) offer unprecedented opportunities for basic and translational research. However, the rapid pace of scientific developments in the field challenges the slower, traditional modes of ethics evaluation. To facilitate responsible research and governance, and ensure public trust, we propose using 'embedded ethics' as a purpose-anchored, dynamic, iterative and integrative approach where ethicists and scientists engage in continuous dialogue to ethically assess ongoing research. We outline a nested benchmarking strategy to periodically evaluate the scientific and ethical status of hSCBEMs within a project, using the human embryo as a reference and weighting criteria along a hierarchy of features that chart embryo-likeness, completeness and the developmental stage modelled. Embedded ethics guides the definition of decision points and ethical boundaries through an iterative assessment of project purpose and ethical and regulatory frameworks, and enables early identification of emerging issues and the co-construction of responsible paths forward.", "authors": [ "Heidi Beate Bentzen", "Maxence Gaillard", "Iftach Nachman", "Daniel Reumann", "Nikolaj Gadegaard", "Laurent David", "Fredrik Lanner", "Naomi Moris", "Vincent Pasque", "Nicolas Rivron", "Berna Sozen", "Rosario Isasi", "Stefan Krauss", "Jesse V Veenvliet" ], "categories": [ "Journal Article", "Review" ], "paper_id": "pubmed:41951755", "abstract_url": "https://pubmed.ncbi.nlm.nih.gov/41951755/", "pdf_url": null, "published_at": "2026-04-08T23:21:00+00:00", "updated_at": "2026-04-08T23:21:00+00:00", "source": "pubmed", "date_label": "Entered", "analysis": null, "tags": [ "评测", "应用", "方法" ], "topics": [ "Benchmark", "Evaluation" ] } ] } ] }