{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,10,28]],"date-time":"2026-10-28T05:06:10Z","timestamp":1793163970321,"version":"4.5.0"},"posted":{"date-parts":[[2026,1,20]]},"group-title":"Bioinformatics","reference-count":154,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,1,20]],"date-time":"2026-01-20T00:00:00Z","timestamp":1768867200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Berkowitz Family Living Laboratory at Harvard Medical School and the Clalit Research Institute"},{"name":"NIH","award":["R01-HD108794"],"award-info":[{"award-number":["R01-HD108794"]}]},{"name":"NSF CAREER","award":["2339524"],"award-info":[{"award-number":["2339524"]}]},{"name":"U.S. DoD","award":["FA8702-15-D-0001"],"award-info":[{"award-number":["FA8702-15-D-0001"]}]},{"name":"ARPA-H Biomedical Data Fabric (BDF) Toolbox Program"},{"DOI":"10.13039\/100018077","name":"Harvard Data Science Initiative","doi-asserted-by":"crossref","id":[{"id":"10.13039\/100018077","id-type":"DOI","asserted-by":"crossref"}]},{"name":"Amazon Faculty Research"},{"name":"Google Research Scholar Program"},{"name":"AstraZeneca Research"},{"name":"Roche Alliance with Distinguished Scientists (ROADS) Program"},{"name":"Sanofi iDEA-iTECH Award"},{"name":"GlaxoSmithKline Award"},{"name":"Boehringer Ingelheim Award"},{"name":"Merck Award"},{"name":"Optum AI Research Collaboration Award"},{"name":"Pfizer Research"},{"name":"Gates Foundation","award":["INV-079038"],"award-info":[{"award-number":["INV-079038"]}]},{"DOI":"10.13039\/100014989","name":"Chan Zuckerberg Initiative","doi-asserted-by":"crossref","id":[{"id":"10.13039\/100014989","id-type":"DOI","asserted-by":"crossref"}]},{"name":"John and Virginia Kaneb Fellowship at Harvard Medical School"},{"name":"Biswas Computational Biology Initiative in partnership with the Milken Institute"},{"name":"Harvard Medical School Dean\u2019s Innovation Fund for the Use of Artificial Intelligence"},{"name":"NIH NCI","award":["U54 CA274502"],"award-info":[{"award-number":["U54 CA274502"]}]},{"name":"NIH NHGRI","award":["R01 HG012351"],"award-info":[{"award-number":["R01 HG012351"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,7,16]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Therapeutic hypotheses can transfer across diseases but their relevance depends on biological context. The same target, perturbation, or treatment can produce different effects across cell types, disease states, genetic backgrounds, and patients. Therapeutic reasoning therefore requires methods that preserve context, test when evidence supports transfer, and identify where context-specific effects limit it. Although AI agents can perform therapeutic analyses, existing systems often fail to preserve biological context over long workflows, verify intermediate computational steps, or reconcile conflicting evidence across datasets and literature. Here, we present Medea, an AI agent for therapeutic reasoning across biological contexts. Medea executes multi-step analyses using biological tools, machine learning models, and literature retrieval while enforcing verification during planning, execution, and evidence synthesis. We evaluate Medea across 5,673 open-ended analyses in three domains: cell type specific therapeutic target nomination in five diseases and 29 cell types, synthetic lethality prediction in 7 cancer cell lines, and immunotherapy response prediction from multimodal patient profiles. Using a previously unpublished epistatic miniarray profiling screen performed under two DNA-damaging treatments, we evaluate Medea on predicting synthetic lethality among 238,046 gene-gene pairs in yeast. Medea accurately predicts these experimentally measured synthetic lethal interactions, indicating that its performance reflects biological relevance rather than information leakage from benchmark datasets. Across these evaluations, Medea improves performance over large language models, reasoning models, biomedical agents, and specialized machine learning models while maintaining low failure rates and calibrated abstention. These results show that verifiable AI agents can perform therapeutic analyses across biological contexts.<\/jats:p>","DOI":"10.64898\/2026.01.16.696667","type":"posted-content","created":{"date-parts":[[2026,1,21]],"date-time":"2026-01-21T02:55:11Z","timestamp":1768964111000},"source":"Crossref","is-referenced-by-count":10,"title":["Medea: An AI agent for therapeutic reasoning across biological contexts"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0009-0001-0891-0477","authenticated-orcid":false,"given":"Pengwei","family":"Sui","sequence":"first","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0223-7485","authenticated-orcid":false,"given":"Michelle M.","family":"Li","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"},{"name":"The Ivan and Francesca Berkowitz Family Living Laboratory Collaboration at Harvard Medical School and Clalit Research Institute, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Brenton P.","family":"Munson","sequence":"additional","affiliation":[{"name":"Department of Medicine, University of California, San Diego, La Jolla, CA, USA"},{"name":"Department of Bioengineering, University of California, San Diego, La Jolla, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shanghua","family":"Gao","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7114-3664","authenticated-orcid":false,"given":"Wanxiang","family":"Shen","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2165-7840","authenticated-orcid":false,"given":"Valentina","family":"Giunchiglia","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"},{"name":"Department of Brain Sciences, Imperial College London, London, UK"},{"name":"Centre for Neuroimaging Sciences, King\u2019s College London, London, UK"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0005-1984-0653","authenticated-orcid":false,"given":"Andrew","family":"Shen","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-5007-4548","authenticated-orcid":false,"given":"Yepeng","family":"Huang","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8120-4456","authenticated-orcid":false,"given":"Zhenglun","family":"Kong","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Katherine","family":"Licon","sequence":"additional","affiliation":[{"name":"Department of Medicine, University of California, San Diego, La Jolla, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1708-8454","authenticated-orcid":false,"given":"Trey","family":"Ideker","sequence":"additional","affiliation":[{"name":"Department of Medicine, University of California, San Diego, La Jolla, CA, USA"},{"name":"Big Data Institute, University of Oxford, Oxford, UK"},{"name":"Department of Computer Science and Engineering, University of California, San Diego, La Jolla, CA, USA"},{"name":"Moores Cancer Center, University of California, San Diego, San Diego, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8530-7228","authenticated-orcid":false,"given":"Marinka","family":"Zitnik","sequence":"additional","affiliation":[{"name":"Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA"},{"name":"The Ivan and Francesca Berkowitz Family Living Laboratory Collaboration at Harvard Medical School and Clalit Research Institute, Boston, MA, USA"},{"name":"Kempner Institute for the Study of Natural and Artificial Intelligence, Harvard University, Allston, MA, USA"},{"name":"Broad Institute of MIT and Harvard, Cambridge, MA, USA"},{"name":"Harvard Data Science Initiative, Cambridge, MA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026072714040346000_2026.01.16.696667v2.1","doi-asserted-by":"crossref","unstructured":"Deshaies, R. J . How multispecific molecules are transforming pharmacotherapy. Nature Reviews Drug Discovery, 1\u201313 (2025).","DOI":"10.1038\/s41573-025-01262-w"},{"key":"2026072714040346000_2026.01.16.696667v2.2","doi-asserted-by":"publisher","DOI":"10.1038\/s41556-025-01626-9"},{"key":"2026072714040346000_2026.01.16.696667v2.3","unstructured":"Alsulami, R. et al. Predicting and interpreting cell-type-specific drug responses in the smalldata regime using inductive priors. Nature Machine Intelligence, 1\u201313 (2026)."},{"key":"2026072714040346000_2026.01.16.696667v2.4","doi-asserted-by":"crossref","first-page":"1025","DOI":"10.1038\/s41586-025-09539-1","article-title":"Proteotoxic stress response drives T cell exhaustion and immune evasion","volume":"647","year":"2025","journal-title":"Nature"},{"key":"2026072714040346000_2026.01.16.696667v2.5","doi-asserted-by":"crossref","unstructured":"Luxenburger, H. , Thimme, R. & Hofmann, M . T cell adaptation in chronic infections and tumors. Cellular & Molecular Immunology, 1\u201317 (2026).","DOI":"10.1038\/s41423-026-01405-y"},{"key":"2026072714040346000_2026.01.16.696667v2.6","doi-asserted-by":"publisher","DOI":"10.1038\/s41568-019-0116-x"},{"key":"2026072714040346000_2026.01.16.696667v2.7","doi-asserted-by":"crossref","first-page":"432","DOI":"10.1038\/s41589-024-01761-8","article-title":"A multiplex single-cell RNA-Seq pharmacotranscriptomics pipeline for drug discovery","volume":"21","year":"2025","journal-title":"Nature Chemical Biology"},{"key":"2026072714040346000_2026.01.16.696667v2.8","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-023-02491-5"},{"key":"2026072714040346000_2026.01.16.696667v2.9","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-024-07606-7"},{"key":"2026072714040346000_2026.01.16.696667v2.10","doi-asserted-by":"crossref","first-page":"9701","DOI":"10.1038\/s41467-025-65783-z","article-title":"AI-powered spatial cell phenomics enhances risk stratification in non-small cell lung cancer","volume":"16","year":"2025","journal-title":"Nature Communications"},{"key":"2026072714040346000_2026.01.16.696667v2.11","unstructured":"Schaffer, L. V. et al. Multimodal cell maps as a foundation for structural and functional genomics. Nature, 1\u201310 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.12","doi-asserted-by":"crossref","first-page":"351","DOI":"10.1038\/s42256-023-00633-5","article-title":"Multimodal data fusion for cancer biomarker discovery with deep learning","volume":"5","year":"2023","journal-title":"Nature machine intelligence"},{"key":"2026072714040346000_2026.01.16.696667v2.13","doi-asserted-by":"crossref","unstructured":"Hingerl, J. C. et al. scooby: modeling multimodal genomic profiles from DNA sequence at single-cell resolution. Nature Methods, 1\u201311 (2025).","DOI":"10.1038\/s41592-025-02854-5"},{"key":"2026072714040346000_2026.01.16.696667v2.14","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2024.09.022"},{"key":"2026072714040346000_2026.01.16.696667v2.15","unstructured":"Gao, S. , et al. Democratizing AI scientists using ToolUniverse. arXiv:2509.23426 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.16","doi-asserted-by":"crossref","unstructured":"Swanson, K. , Wu, W. , Bulaong, N. L. , Pak, J. E. & Zou, J . The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies. Nature, 1\u20133 (2025).","DOI":"10.1101\/2024.11.11.623004"},{"key":"2026072714040346000_2026.01.16.696667v2.17","doi-asserted-by":"crossref","first-page":"9377","DOI":"10.1038\/s41467-025-64430-x","article-title":"Agentmd: Empowering language agents for risk prediction with large-scale clinical tool learning","volume":"16","year":"2025","journal-title":"Nature Communications"},{"key":"2026072714040346000_2026.01.16.696667v2.18","doi-asserted-by":"crossref","first-page":"82","DOI":"10.1038\/s41592-024-02525-x","article-title":"Evaluation of large language models for discovery of gene set function","volume":"22","year":"2025","journal-title":"Nature Methods"},{"key":"2026072714040346000_2026.01.16.696667v2.19","unstructured":"Wang, Z. et al. GeneAgent: self-verification language agent for gene-set analysis using domain databases. Nature Methods, 1\u20139 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.20","doi-asserted-by":"crossref","unstructured":"Alber, S. et al. CellVoyager: Ai compbio agent generates new insights by autonomously analyzing biological data. Nature Methods, 1\u201311 (2026).","DOI":"10.1101\/2025.06.03.657517"},{"key":"2026072714040346000_2026.01.16.696667v2.21","doi-asserted-by":"crossref","unstructured":"Huang, K. , et al. Biomni: A general-purpose biomedical AI agent. biorxiv (2025).","DOI":"10.1101\/2025.05.30.656746"},{"key":"2026072714040346000_2026.01.16.696667v2.22","doi-asserted-by":"crossref","unstructured":"Qu, Y. et al. CRISPR-GPT for agentic automation of gene-editing experiments. Nature Biomedical Engineering, 1\u201314 (2025).","DOI":"10.1101\/2024.04.25.591003"},{"key":"2026072714040346000_2026.01.16.696667v2.23","doi-asserted-by":"crossref","unstructured":"Luo, E. , et al. Benchmarking AI scientists in omics data-driven biological research. arXiv:2505.08341 (2025).","DOI":"10.1093\/bioinformatics\/btag227"},{"key":"2026072714040346000_2026.01.16.696667v2.24","unstructured":"Costanzo, M. et al. A global genetic interaction network maps a wiring diagram of cellular function. Science 353, aaf1420 (2016)."},{"key":"2026072714040346000_2026.01.16.696667v2.25","doi-asserted-by":"publisher","DOI":"10.1038\/nrc1547"},{"key":"2026072714040346000_2026.01.16.696667v2.26","doi-asserted-by":"crossref","first-page":"1546","DOI":"10.1038\/s41592-024-02341-3","article-title":"Contextual AI models for single-cell protein biology","volume":"21","year":"2024","journal-title":"Nature Methods"},{"key":"2026072714040346000_2026.01.16.696667v2.27","doi-asserted-by":"crossref","unstructured":"Pearce, J. D. et al. TranscriptFormer: A generative cell atlas across 1.5 billion years of evolution. Science, eaec8514 (2026).","DOI":"10.1126\/science.aec8514"},{"key":"2026072714040346000_2026.01.16.696667v2.28","doi-asserted-by":"crossref","unstructured":"Shen, W. et al. Generalizable AI predicts immunotherapy outcomes across cancers and treatments. Nature Medicine, 2026\u201305 (2026).","DOI":"10.1038\/s41591-026-04502-7"},{"key":"2026072714040346000_2026.01.16.696667v2.29","doi-asserted-by":"publisher","DOI":"10.1038\/ng.3259"},{"key":"2026072714040346000_2026.01.16.696667v2.30","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkac240"},{"key":"2026072714040346000_2026.01.16.696667v2.31","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkae1128"},{"key":"2026072714040346000_2026.01.16.696667v2.32","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkx1132"},{"key":"2026072714040346000_2026.01.16.696667v2.33","doi-asserted-by":"crossref","first-page":"417","DOI":"10.1016\/j.cels.2015.12.004","article-title":"The molecular signatures database hallmark gene set collection","volume":"1","year":"2015","journal-title":"Cell systems"},{"key":"2026072714040346000_2026.01.16.696667v2.34","unstructured":"Aleksander, S. A. et al. The gene ontology knowledgebase in 2023. Genetics 224, iyad031 (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.35","doi-asserted-by":"publisher","DOI":"10.1038\/s41568-024-00763-x"},{"key":"2026072714040346000_2026.01.16.696667v2.36","doi-asserted-by":"publisher","DOI":"10.1126\/science.1260419"},{"key":"2026072714040346000_2026.01.16.696667v2.37","unstructured":"Thul, P. J. et al. A subcellular map of the human proteome. Science 356, eaal3321 (2017)."},{"key":"2026072714040346000_2026.01.16.696667v2.38","unstructured":"Karlsson, M. et al. A single\u2013cell type transcriptomics map of human tissues. Science advances 7, eabh2169 (2021)."},{"key":"2026072714040346000_2026.01.16.696667v2.39","unstructured":"Kinney, R. , et al. The semantic scholar open data platform. arXiv:2301.10140 (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.40","unstructured":"Priem, J. , Piwowar, H. & Orr, R . OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts. arXiv:2205.01833 (2022)."},{"key":"2026072714040346000_2026.01.16.696667v2.41","unstructured":"Asai, A. et al. Openscholar: Synthesizing scientific literature with retrieval-augmented lms. arXiv:2411.14199 (2024)."},{"key":"2026072714040346000_2026.01.16.696667v2.42","first-page":"38975","article-title":"Planbench: An extensible benchmark for evaluating large language models on planning and reasoning about change","volume":"36","year":"2023","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026072714040346000_2026.01.16.696667v2.43","unstructured":"Wang, X. et al. Self-consistency improves chain of thought reasoning in language models. arXiv:2203.11171 (2022)."},{"key":"2026072714040346000_2026.01.16.696667v2.44","unstructured":"Zhang, W. , et al. Deep research: A survey of autonomous research agents. arXiv:2508.12752 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.45","doi-asserted-by":"crossref","unstructured":"Wei, H. , et al. Plangenllms: A modern survey of llm planning capabilities. arXiv:2502.11221 (2025).","DOI":"10.18653\/v1\/2025.acl-long.958"},{"key":"2026072714040346000_2026.01.16.696667v2.46","unstructured":"Longpre, S. , et al. The data provenance initiative: A large scale audit of dataset licensing & attribution in ai (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.47","doi-asserted-by":"crossref","unstructured":"Chen, J. , Saha, S. & Bansal, M. Reconcile: Round-table conference improves reasoning via consensus among diverse llms in Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (2024), 7066\u20137085.","DOI":"10.18653\/v1\/2024.acl-long.381"},{"key":"2026072714040346000_2026.01.16.696667v2.48","doi-asserted-by":"publisher","DOI":"10.1038\/nrd1202"},{"key":"2026072714040346000_2026.01.16.696667v2.49","unstructured":"Laman Trip, D. S. , et al. A tissue-specific atlas of protein\u2013protein associations enables prioritization of candidate disease genes. Nature Biotechnology, 1\u201314 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.50","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-022-31543-6"},{"key":"2026072714040346000_2026.01.16.696667v2.51","doi-asserted-by":"crossref","first-page":"6494","DOI":"10.1038\/s41467-022-34277-7","article-title":"Deep transfer learning of cancer drug responses by integrating bulk and single-cell RNA-seq data","volume":"13","year":"2022","journal-title":"Nature Communications"},{"key":"2026072714040346000_2026.01.16.696667v2.52","doi-asserted-by":"crossref","first-page":"1853","DOI":"10.1038\/s41467-024-46089-y","article-title":"Drug target prediction through deep learning functional representation of gene signatures","volume":"15","year":"2024","journal-title":"Nature communications"},{"key":"2026072714040346000_2026.01.16.696667v2.53","doi-asserted-by":"crossref","first-page":"e178499","DOI":"10.1172\/jci.insight.178499","article-title":"Single-cell RNA-Seq analysis reveals cell subsets and gene signatures associated with rheumatoid arthritis disease activity","volume":"9","year":"2024","journal-title":"JCI insight"},{"key":"2026072714040346000_2026.01.16.696667v2.54","doi-asserted-by":"crossref","first-page":"284","DOI":"10.1038\/s42255-022-00531-x","article-title":"Single-cell multi-omics analysis of human pancreatic islets reveals novel cellular states in type 1 diabetes","volume":"4","year":"2022","journal-title":"Nature Metabolism"},{"key":"2026072714040346000_2026.01.16.696667v2.55","unstructured":"Pranzatelli, T. J. et al. GZMK+ CD8+ T cells target a specific Acinar cell type in Sj\u00f6gren\u2019s disease. Research Square, rs\u20133 (2024)."},{"key":"2026072714040346000_2026.01.16.696667v2.56","doi-asserted-by":"crossref","first-page":"1049","DOI":"10.1038\/s42003-021-02562-8","article-title":"Identification of distinct tumor cell populations and key genetic mechanisms through single cell sequencing in hepatoblastoma","volume":"4","year":"2021","journal-title":"Communications Biology"},{"key":"2026072714040346000_2026.01.16.696667v2.57","doi-asserted-by":"publisher","DOI":"10.1158\/2643-3230.BCD-21-0075"},{"key":"2026072714040346000_2026.01.16.696667v2.58","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-023-01609-2"},{"key":"2026072714040346000_2026.01.16.696667v2.59","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkad1004"},{"key":"2026072714040346000_2026.01.16.696667v2.60","first-page":"1","article-title":"Statistical comparisons of classifiers over multiple data sets","volume":"7","year":"2006","journal-title":"Journal of Machine Learning Research"},{"key":"2026072714040346000_2026.01.16.696667v2.61","unstructured":"Chern, I. , et al. FacTool: Factuality Detection in Generative AI\u2013A Tool Augmented Framework for Multi-Task and Multi-Domain Scenarios. arXiv:2307.13528 (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.62","unstructured":"Tian, K. , Mitchell, E. , Yao, H. , Manning, C. D. & Finn, C. Fine-tuning language models for factuality in The Twelfth International Conference on Learning Representations (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.63","unstructured":"Gou, Z. , et al. Critic: Large language models can self-correct with tool-interactive critiquing. arXiv:2305.11738 (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.64","unstructured":"Cheng, Q. , et al. Can AI assistants know what they don\u2019t know? arXiv:2401.13275 (2024)."},{"key":"2026072714040346000_2026.01.16.696667v2.65","doi-asserted-by":"crossref","unstructured":"Brahman, F. et al. The art of saying no: Contextual noncompliance in language models, 2024. URL https:\/\/arxiv.org\/abs\/240712043.","DOI":"10.52202\/079017-1573"},{"key":"2026072714040346000_2026.01.16.696667v2.66","doi-asserted-by":"crossref","first-page":"529","DOI":"10.1162\/tacl_a_00754","article-title":"Know your limits: A survey of abstention in large language models","volume":"13","year":"2025","journal-title":"Transactions of the Association for Computational Linguistics"},{"key":"2026072714040346000_2026.01.16.696667v2.67","unstructured":"Liu, X. et al. Agentbench: Evaluating llms as agents. arXiv:2308.03688 (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.68","first-page":"23813","article-title":"Swiftsage: A generative agent with fast and slow thinking for complex interactive tasks","volume":"36","year":"2023","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026072714040346000_2026.01.16.696667v2.69","doi-asserted-by":"publisher","DOI":"10.1038\/44385"},{"key":"2026072714040346000_2026.01.16.696667v2.70","doi-asserted-by":"publisher","DOI":"10.1038\/nri3567"},{"key":"2026072714040346000_2026.01.16.696667v2.71","doi-asserted-by":"crossref","unstructured":"Shi, W. et al. Trusting your evidence: Hallucinate less with context-aware decoding, 783\u2013 791 (2024).","DOI":"10.18653\/v1\/2024.naacl-short.69"},{"key":"2026072714040346000_2026.01.16.696667v2.72","doi-asserted-by":"crossref","unstructured":"Zhou, W. , Zhang, S. , Poon, H. & Chen, M. Context-faithful prompting for large language models. arXiv:2303.11315 (2023).","DOI":"10.18653\/v1\/2023.findings-emnlp.968"},{"key":"2026072714040346000_2026.01.16.696667v2.73","doi-asserted-by":"publisher","DOI":"10.1016\/j.immuni.2024.11.004"},{"key":"2026072714040346000_2026.01.16.696667v2.74","doi-asserted-by":"crossref","first-page":"353","DOI":"10.3389\/fimmu.2019.00353","article-title":"Effector functions of CD4+ T cells at the site of local autoimmune inflammation\u2014lessons from rheumatoid arthritis","volume":"10","year":"2019","journal-title":"Frontiers in Immunology"},{"key":"2026072714040346000_2026.01.16.696667v2.75","doi-asserted-by":"crossref","first-page":"319","DOI":"10.1038\/s41467-022-35264-8","article-title":"Cytotoxic CD8+ T cells target citrullinated antigens in rheumatoid arthritis","volume":"14","year":"2023","journal-title":"Nature Communications"},{"key":"2026072714040346000_2026.01.16.696667v2.76","doi-asserted-by":"crossref","unstructured":"Xiang, N. , et al. Single-cell transcriptome profiling reveals immune and stromal cell heterogeneity in primary Sj\u00f6gren\u2019s syndrome. iScience 26 (2023).","DOI":"10.1016\/j.isci.2023.107943"},{"key":"2026072714040346000_2026.01.16.696667v2.77","doi-asserted-by":"publisher","DOI":"10.1093\/rheumatology\/41.7.819"},{"key":"2026072714040346000_2026.01.16.696667v2.78","doi-asserted-by":"crossref","first-page":"421","DOI":"10.1002\/art.42723","article-title":"Increased diagnostic accuracy of the labial gland biopsy in primary sj\u00f6gren syndrome when multiple histopathological features are included","volume":"76","year":"2024","journal-title":"Arthritis & Rheumatology"},{"key":"2026072714040346000_2026.01.16.696667v2.79","doi-asserted-by":"publisher","DOI":"10.1016\/j.imlet.2023.06.005"},{"key":"2026072714040346000_2026.01.16.696667v2.80","doi-asserted-by":"crossref","first-page":"316","DOI":"10.1038\/cmi.2015.104","article-title":"Hepatic macrophages in homeostasis and liver diseases: from pathogenesis to novel therapeutic strategies","volume":"13","year":"2016","journal-title":"Cellular & Molecular Immunology"},{"key":"2026072714040346000_2026.01.16.696667v2.81","doi-asserted-by":"publisher","DOI":"10.1038\/nri3175"},{"key":"2026072714040346000_2026.01.16.696667v2.82","doi-asserted-by":"crossref","unstructured":"Monti, M. et al. Plasmacytoid dendritic cells at the forefront of anti-cancer immunity: rewiring strategies for tumor microenvironment remodeling. Journal of Experimental & Clinical Cancer Research 43, 196 (2024).","DOI":"10.1186\/s13046-024-03121-9"},{"key":"2026072714040346000_2026.01.16.696667v2.83","doi-asserted-by":"publisher","DOI":"10.1038\/nrg.2017.47"},{"key":"2026072714040346000_2026.01.16.696667v2.84","doi-asserted-by":"crossref","unstructured":"Gon\u00e7alves, E. , Ryan, C. J. & Adams, D. J . Synthetic lethality in cancer drug discovery: challenges and opportunities. Nature Reviews Drug Discovery, 1\u201317 (2025).","DOI":"10.1038\/s41573-025-01273-7"},{"key":"2026072714040346000_2026.01.16.696667v2.85","doi-asserted-by":"crossref","first-page":"154","DOI":"10.1038\/s41588-024-01971-9","article-title":"A multilineage screen identifies actionable synthetic lethal interactions in human cancers","volume":"57","year":"2025","journal-title":"Nature Genetics"},{"key":"2026072714040346000_2026.01.16.696667v2.86","doi-asserted-by":"crossref","first-page":"1661","DOI":"10.1038\/s41467-021-21898-7","article-title":"Integrated cross-study datasets of genetic dependencies in cancer","volume":"12","year":"2021","journal-title":"Nature Communications"},{"key":"2026072714040346000_2026.01.16.696667v2.87","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkw377"},{"key":"2026072714040346000_2026.01.16.696667v2.88","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkaa1024"},{"key":"2026072714040346000_2026.01.16.696667v2.89","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkh052"},{"key":"2026072714040346000_2026.01.16.696667v2.90","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btr260"},{"key":"2026072714040346000_2026.01.16.696667v2.91","doi-asserted-by":"crossref","unstructured":"Costanzo, M. et al. Environmental robustness of the global yeast genetic interaction network. Science 372, eabf8424 (2021).","DOI":"10.1126\/science.abf8424"},{"key":"2026072714040346000_2026.01.16.696667v2.92","doi-asserted-by":"publisher","DOI":"10.1093\/genetics\/iyae049"},{"key":"2026072714040346000_2026.01.16.696667v2.93","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkr1029"},{"key":"2026072714040346000_2026.01.16.696667v2.94","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/26.1.73"},{"key":"2026072714040346000_2026.01.16.696667v2.95","doi-asserted-by":"crossref","unstructured":"Carme, P. , Rutherford, K. , B\u00e4hler, J. , Mata, J. & Wood, V. PomBase in 2026: Expanding Knowledge, Modelling Connections. Genetics, iyag001 (2026).","DOI":"10.1093\/genetics\/iyag001"},{"key":"2026072714040346000_2026.01.16.696667v2.96","doi-asserted-by":"publisher","DOI":"10.1093\/NAR\/GKAC1000"},{"key":"2026072714040346000_2026.01.16.696667v2.97","doi-asserted-by":"publisher","DOI":"10.1038\/s41422-020-0343-4"},{"key":"2026072714040346000_2026.01.16.696667v2.98","doi-asserted-by":"crossref","first-page":"320","DOI":"10.1038\/s41392-023-01522-4","article-title":"Immune checkpoint therapy for solid tumours: clinical dilemmas and future trends","volume":"8","year":"2023","journal-title":"Signal Transduction and Targeted Therapy"},{"key":"2026072714040346000_2026.01.16.696667v2.99","doi-asserted-by":"crossref","first-page":"3044","DOI":"10.1093\/annonc\/mdx518","article-title":"Atezolizumab in platinum-treated locally advanced or metastatic urothelial carcinoma: post-progression outcomes from the phase II IMvigor210 study","volume":"28","year":"2017","journal-title":"Annals of Oncology"},{"key":"2026072714040346000_2026.01.16.696667v2.100","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-019-0654-5"},{"key":"2026072714040346000_2026.01.16.696667v2.101","doi-asserted-by":"crossref","first-page":"274","DOI":"10.1038\/s41392-024-01979-x","article-title":"Cold and hot tumors: from molecular mechanisms to targeted therapy","volume":"9","year":"2024","journal-title":"Signal Transduction and Targeted Therapy"},{"key":"2026072714040346000_2026.01.16.696667v2.102","doi-asserted-by":"crossref","first-page":"5034","DOI":"10.1158\/1078-0432.CCR-24-1883","article-title":"Plasma ctDNA as a treatment response biomarker in metastatic cancers: evaluation by the RECIST working group","volume":"30","year":"2024","journal-title":"Clinical Cancer Research"},{"key":"2026072714040346000_2026.01.16.696667v2.103","doi-asserted-by":"crossref","first-page":"725","DOI":"10.1038\/s41571-024-00932-9","article-title":"Tumour mutational burden: clinical utility, challenges and emerging improvements","volume":"21","year":"2024","journal-title":"Nature Reviews Clinical Oncology"},{"key":"2026072714040346000_2026.01.16.696667v2.104","doi-asserted-by":"crossref","unstructured":"Epperson, W. , et al. Interactive debugging and steering of multi-agent ai systems in Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (2025), 1\u2013 15.","DOI":"10.1145\/3706598.3713581"},{"key":"2026072714040346000_2026.01.16.696667v2.105","unstructured":"Zhang, Z. et al. Multi-Agent Taskforce Collaboration: Self-Correction of Compounding Errors in Long-Form Literature Review Generation. arXiv:2508.04306 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.106","first-page":"75749","article-title":"GTA: a benchmark for general tool agents","volume":"37","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026072714040346000_2026.01.16.696667v2.107","unstructured":"Vuddanti, S. V. et al. PALADIN: Self-Correcting Language Model Agents to Cure ToolFailure Cases. arXiv:2509.25238 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.108","doi-asserted-by":"crossref","unstructured":"Errica, F. , Sanvito, D. , Siracusano, G. & Bifulco, R. What did i do wrong? quantifying LLMs\u2019 sensitivity and consistency to prompt engineering in Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers) (2025), 1543\u20131558.","DOI":"10.18653\/v1\/2025.naacl-long.73"},{"key":"2026072714040346000_2026.01.16.696667v2.109","first-page":"46595","article-title":"Judging LLM-as-a-judge with MT-Bench and chatbot arena","volume":"36","year":"2023","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026072714040346000_2026.01.16.696667v2.110","unstructured":"Kim, E. , Garg, A. , Peng, K. & Garg, N . Correlated Errors in Large Language Models. International Conference on Machine Learning (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.111","doi-asserted-by":"crossref","unstructured":"Liu, X. et al. PiLSL: pairwise interaction learning-based graph neural network for synthetic lethality prediction in human cancers. Bioinformatics 38, ii106\u2013ii112 (2022).","DOI":"10.1093\/bioinformatics\/btac476"},{"key":"2026072714040346000_2026.01.16.696667v2.112","doi-asserted-by":"crossref","unstructured":"Fasolino, M. et al. Multiomics single-cell analysis of human pancreatic islets reveals novel cellular states in health and type 1 diabetes. BioRxiv, 2021\u201301 (2021).","DOI":"10.1101\/2021.01.28.428598"},{"key":"2026072714040346000_2026.01.16.696667v2.113","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkae1142"},{"key":"2026072714040346000_2026.01.16.696667v2.114","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkae1059"},{"key":"2026072714040346000_2026.01.16.696667v2.115","unstructured":"Hagberg, A. , Swart, P. J. & Schult, D. A . Exploring network structure, dynamics, and func tion using NetworkX tech. rep. (Los Alamos National Laboratory (LANL), Los Alamos, NM (United States), 2008)."},{"key":"2026072714040346000_2026.01.16.696667v2.116","doi-asserted-by":"publisher","DOI":"10.1002\/PRO.3978"},{"key":"2026072714040346000_2026.01.16.696667v2.117","doi-asserted-by":"publisher","DOI":"10.1016\/S0014-5793(01)03293-8"},{"key":"2026072714040346000_2026.01.16.696667v2.118","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/30.1.31"},{"key":"2026072714040346000_2026.01.16.696667v2.119","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkad1059"},{"key":"2026072714040346000_2026.01.16.696667v2.120","doi-asserted-by":"publisher","DOI":"10.1038\/75556"},{"key":"2026072714040346000_2026.01.16.696667v2.121","doi-asserted-by":"publisher","DOI":"10.1186\/1471-2105-14-128"},{"key":"2026072714040346000_2026.01.16.696667v2.122","doi-asserted-by":"publisher","DOI":"10.1213\/ANE.0000000000002864"},{"key":"2026072714040346000_2026.01.16.696667v2.123","doi-asserted-by":"publisher","DOI":"10.2307\/2346101"},{"key":"2026072714040346000_2026.01.16.696667v2.124","doi-asserted-by":"publisher","DOI":"10.1186\/s13059-021-02540-7"},{"key":"2026072714040346000_2026.01.16.696667v2.125","first-page":"46534","article-title":"Self-refine: Iterative refinement with self-feedback","volume":"36","year":"2023","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026072714040346000_2026.01.16.696667v2.126","unstructured":"Jiang, S. , Wang, Y. & Wang, Y . Selfevolve: A code evolution framework via large language models. arXiv:2306.02907 (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.127","doi-asserted-by":"crossref","unstructured":"Chen, J. , et al. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. arXiv:2402.03216 (2024).","DOI":"10.18653\/v1\/2024.findings-acl.137"},{"key":"2026072714040346000_2026.01.16.696667v2.128","unstructured":"Google. Gemini 2.0 Flash Model Card tech. rep. (Google, 2024). https:\/\/modelcards.withgoogle.com\/assets\/documents\/gemini-2-flash.pdf."},{"key":"2026072714040346000_2026.01.16.696667v2.129","unstructured":"Zhang, B. , et al. OpenAI o3-mini System Card in (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.130","unstructured":"Hurst, A. , et al. Gpt-4o system card. arXiv:2410.21276 (2024)."},{"key":"2026072714040346000_2026.01.16.696667v2.131","unstructured":"Anthropic. Claude 3.7 Sonnet System Card in (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.132","unstructured":"Liu, Z. et al. Agentlite: A lightweight library for building and advancing task-oriented llm agent system. arXiv:2402.15538 (2024)."},{"key":"2026072714040346000_2026.01.16.696667v2.133","unstructured":"Yao, S. , et al. React: Synergizing reasoning and acting in language models in International Conference on Learning Representations (ICLR) (2023)."},{"key":"2026072714040346000_2026.01.16.696667v2.134","doi-asserted-by":"crossref","unstructured":"Megill, C. et al. CELLxGENE: a performant, scalable exploration platform for high dimensional sparse matrices. bioRxiv, 2021\u201304 (2021).","DOI":"10.1101\/2021.04.05.438318"},{"key":"2026072714040346000_2026.01.16.696667v2.135","doi-asserted-by":"publisher","DOI":"10.1186\/s13059-017-1382-0"},{"key":"2026072714040346000_2026.01.16.696667v2.136","doi-asserted-by":"publisher","DOI":"10.15252\/msb.20188746"},{"key":"2026072714040346000_2026.01.16.696667v2.137","doi-asserted-by":"publisher","DOI":"10.1016\/J.CELL.2021.04.048"},{"key":"2026072714040346000_2026.01.16.696667v2.138","doi-asserted-by":"publisher","DOI":"10.1038\/ng.3314"},{"key":"2026072714040346000_2026.01.16.696667v2.139","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2017.06.010"},{"key":"2026072714040346000_2026.01.16.696667v2.140","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-019-1103-9"},{"key":"2026072714040346000_2026.01.16.696667v2.141","doi-asserted-by":"crossref","unstructured":"Licon, K. et al. in Yeast Systems Biology: Methods and Protocols 73\u201385 (Springer, 2019).","DOI":"10.1007\/978-1-4939-9736-7_4"},{"key":"2026072714040346000_2026.01.16.696667v2.142","doi-asserted-by":"crossref","first-page":"573","DOI":"10.1038\/nmeth.4225","article-title":"Combinatorial CRISPR\u2013Cas9 screens for de novo mapping of genetic interactions","volume":"14","year":"2017","journal-title":"Nature methods"},{"key":"2026072714040346000_2026.01.16.696667v2.143","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-018-0312-8"},{"key":"2026072714040346000_2026.01.16.696667v2.144","first-page":"30039","article-title":"Alpacafarm: A simulation framework for methods that learn from human feedback","volume":"36","year":"2023","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2026072714040346000_2026.01.16.696667v2.145","doi-asserted-by":"crossref","unstructured":"Kamath, A. , Jia, R. & Liang, P. Selective question answering under domain shift. arXiv preprint arXiv:2006.09462 (2020).","DOI":"10.18653\/v1\/2020.acl-main.503"},{"key":"2026072714040346000_2026.01.16.696667v2.146","doi-asserted-by":"crossref","unstructured":"Cole, J. , et al. Selectively answering ambiguous questions in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (2023), 530\u2013543.","DOI":"10.18653\/v1\/2023.emnlp-main.35"},{"key":"2026072714040346000_2026.01.16.696667v2.147","unstructured":"Si, C. , et al. Prompting gpt-3 to be reliable. arXiv preprint arXiv:2210.09150 (2022)."},{"key":"2026072714040346000_2026.01.16.696667v2.148","unstructured":"Golovneva, O. , et al. Roscoe: A suite of metrics for scoring step-by-step reasoning. arXiv preprint arXiv:2212.07919 (2022)."},{"key":"2026072714040346000_2026.01.16.696667v2.149","doi-asserted-by":"crossref","unstructured":"Wadden, D. et al. Fact or Fiction: Verifying Scientific Claims in Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) (eds Webber, B., Cohn, T., He, Y. & Liu, Y.) (Association for Computational Linguistics, Online, Nov. 2020), 7534\u20137550. https:\/\/aclanthology.org\/2020.emnlp-main.609\/.","DOI":"10.18653\/v1\/2020.emnlp-main.609"},{"key":"2026072714040346000_2026.01.16.696667v2.150","unstructured":"Jaech, A. et al. Openai o1 system card. arXiv:2412.16720 (2024)."},{"key":"2026072714040346000_2026.01.16.696667v2.151","unstructured":"Guo, D. , et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv:2501.12948 (2025)."},{"key":"2026072714040346000_2026.01.16.696667v2.152","unstructured":"El-Yaniv, R. et al. On the Foundations of Noise-free Selective Classification. Journal of Machine Learning Research 11 (2010)."},{"key":"2026072714040346000_2026.01.16.696667v2.153","unstructured":"Geifman, Y. & El-Yaniv, R . Selective classification for deep neural networks. Advances in neural information processing systems 30 (2017)."},{"key":"2026072714040346000_2026.01.16.696667v2.154","doi-asserted-by":"crossref","unstructured":"Seabold, S. & Perktold, J. Statsmodels: Econometric and statistical modeling with python in 9th Python in Science Conference (2010).","DOI":"10.25080\/Majora-92bf1922-011"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.01.16.696667","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,27]],"date-time":"2026-07-27T21:04:25Z","timestamp":1785186265000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.64898\/2026.01.16.696667"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,20]]},"references-count":154,"URL":"https:\/\/doi.org\/10.64898\/2026.01.16.696667","relation":{},"subject":[],"published":{"date-parts":[[2026,1,20]]},"subtype":"preprint"}}