[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81750-en":3,"doc-seo-81750-105":28,"detail-sidebar-cat-0-en-105":89},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":13,"seo_description":14,"update_tm":26,"read_time":27},81750,4398048949847,"Eliana","https://ap-avatar.wpscdn.com/avatar/400002536579ef2da7f?_k=1778318612642679267",7,"Healthcare","RareDxR1 Autonomous Medical Reasoning for Rare Disease Diagnosis Beyond Human Annotation","RareDxR1 addresses rare disease differential diagnosis, a complex clinical reasoning task requiring accurate phenotype identification from unstructured notes within a large hypothesis space. It replaces pipeline-style phenotype extraction and retrieval-augmented generation that lose information through ontologies and retrieval bottlenecks by using an end-to-end reasoning-centric large language model. The model deep-internalizes fragmented rare disease knowledge into parameters, then enhances diagnostic trajectories with Reflection-Enhanced Reasoning Sampling learned from failures without human annotation, supported by dual-level curriculum reinforcement learning.","RareDxR1: Autonomous Medical Reasoning for Rare Disease Diagnosis Beyond Human Annotation  \nDeyang Jiang 1 ,2 ,∗ , Haoran Wu 1 ,∗ , Ziyi Wang 1 , Yiming Rong 1 , Yunlong Zhao 1 , Ye Jin3 , and Bo Xu 1 ,2  \n1The Key Laboratory of Cognition and Decision Intelligence for Complex Systems,  \nInstitute of Automation, Chinese Academy of Sciences, Beijing, China  \n2 School of Future Technology, University of Chinese Academy of Sciences, Beijing, China  \n3Peking Union Medical College Hospital, Beijing, China  \narXiv :2607 .00 147v 1 [ cs .AI] 30 Jun 2026  \nAbstract—Rare disease differential diagnosis is a critical yet arduous clinical task, requiring physicians to identify precise phenotypes from complex, unstructured patient symptoms and execute intricate reasoning within a vast search space. However, existing AI approaches typically rely on pipeline-based phenotype extraction or retrieval-augmented generation, which suffer from critical information loss due to predefined ontologies, retrieval bottlenecks, and a lack of diagnostic logic. To address these challenges, we introduce RareDxR1, an end-to-end reasoningcentric large language model designed for open-domain rare disease diagnosis directly from unstructured clinical notes. We design a progressive end-to-end training framework by synergizing knowledge internalization with autonomous evolutionary learning, thereby bypassing reliance on structured phenotypesand closed-set decision-making. To overcome the limitations of RAG and phenotype restriction, we enabled the deep internalization of fragmented rare disease knowledge directly into the model’s parameters. Moreover, to bridge the gap between model generation and expert reasoning, we propose ReflectionEnhanced Reasoning Sampling (RERS), a strategy that synthesizes expert-level diagnostic trajectories by learning from failures without human annotation. Additionally, we propose a dual-level curriculum reinforcement learning approach for gradually mastering rare disease diagnosis. Experimental results demonstrate that RareDxR1 achieves a state-of-the-art accuracy across different benchmarks, marking a significant breakthrough in open-domain rare disease diagnosis. Our code and dataset will be publicly available.1  \nIndex Terms—Rare Disease Diagnosis, Large Language Models, Reinforcement Learning, Chain-of-Thought, Medical Reasoning  \n∗ Equal contribution. Corresponding author: Bo Xu ([xubo@ia.ac.cn](xubo@ia.ac.cn)).  \nThe project is supported by the National High Level Hospital Clinical Research Funding (2025-PUMCH-C-006), and the National Natural Science Foundation of China (No. 62506358) .  \nAccepted to the IEEE International Conference on Multimedia and Expo (ICME) 2026 . © 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.  \nThe project is supported by the National High Level Hospital Clinical Research Funding (2025-PUMCH-C-006), and the National Natural Science Foundation of China (No. 62506358) .  \n1[https://github.com/jdy18/RareDxR1](https://github.com/jdy18/RareDxR1)  \nFig. 1. Top-10 Recall across all rare disease categories in RareArena-Test.  \nI. INTRODUCTION  \nRare diseases, affecting less than 1/2000 of the population, encompass 7,000–10,000 types globally and impact approximately 300 million people [1], [2] . Rare Disease Diagnosis (RDD)—the process of identifying specific diseases from undifferentiated clinical manifestations—is notoriously difficult. In clinical practice, this task requires a dual-process capability: first, physicians must meticulously parse unstructured patient narratives to identify key phenotypes; second, they must perform complex differential diagnosis, weighing conflicting","cbCaiphIimxojP9T","https://ap.wps.com/l/cbCaiphIimxojP9T","pdf",941452,1,"English","en",105,"# Introduction\n## Problem: rare disease differential diagnosis\n## Limitations of ontology/RAG-based approaches\n## Knowledge and reasoning gaps in medical LLMs\n## Proposed approach: RareDxR1","[{\"question\":\"What core problem does RareDxR1 target in rare disease diagnosis?\",\"answer\":\"It targets the difficulty of performing differential diagnosis by deriving precise phenotypes from unstructured patient symptoms and executing multi-step reasoning over a large search space.\"},{\"question\":\"How does RareDxR1 differ from pipeline phenotype extraction or RAG methods?\",\"answer\":\"It performs end-to-end diagnostic inference directly from unstructured clinical notes, avoiding rigid phenotype extraction and reducing information loss caused by predefined ontologies and retrieval bottlenecks.\"},{\"question\":\"How does RareDxR1 improve reasoning without human annotation?\",\"answer\":\"It introduces Reflection-Enhanced Reasoning Sampling (RERS) to synthesize expert-level diagnostic trajectories by learning from failures, without relying on human annotation.\"}]",1784175823,18,{"code":4,"msg":29,"data":30},"ok",{"site_id":23,"language":22,"slug":31,"title":13,"keywords":32,"description":14,"schema_data":33,"social_meta":84,"head_meta":86,"extra_data":88,"updated_unix":26},"raredxr1-autonomous-medical-reasoning-for-rare-disease-diagnosis-beyond-human-annotation","",{"@graph":34,"@context":83},[35,52,66],{"@type":36,"itemListElement":37},"BreadcrumbList",[38,42,46,49],{"item":39,"name":40,"@type":41,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":43,"name":44,"@type":41,"position":45},"https://docshare.wps.com/document/","Document",2,{"item":47,"name":12,"@type":41,"position":48},"https://docshare.wps.com/document/healthcare/",3,{"item":50,"name":13,"@type":41,"position":51},"https://docshare.wps.com/document/raredxr1-autonomous-medical-reasoning-for-rare-disease-diagnosis-beyond-human-annotation/81750/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":22,"description":14,"dateModified":60,"datePublished":60,"encodingFormat":59,"isAccessibleForFree":61,"interactionStatistic":62},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":39,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":63,"interactionType":64,"userInteractionCount":4},"InteractionCounter",{"@type":65},"ViewAction",{"@type":67,"mainEntity":68},"FAQPage",[69,75,79],{"name":70,"@type":71,"acceptedAnswer":72},"What core problem does RareDxR1 target in rare disease diagnosis?","Question",{"text":73,"@type":74},"It targets the difficulty of performing differential diagnosis by deriving precise phenotypes from unstructured patient symptoms and executing multi-step reasoning over a large search space.","Answer",{"name":76,"@type":71,"acceptedAnswer":77},"How does RareDxR1 differ from pipeline phenotype extraction or RAG methods?",{"text":78,"@type":74},"It performs end-to-end diagnostic inference directly from unstructured clinical notes, avoiding rigid phenotype extraction and reducing information loss caused by predefined ontologies and retrieval bottlenecks.",{"name":80,"@type":71,"acceptedAnswer":81},"How does RareDxR1 improve reasoning without human annotation?",{"text":82,"@type":74},"It introduces Reflection-Enhanced Reasoning Sampling (RERS) to synthesize expert-level diagnostic trajectories by learning from failures, without relying on human annotation.","https://schema.org",{"og:url":50,"og:type":85,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":87,"canonical":50},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":90},[91,95,99,103,108,113,116,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":44,"category_name":92,"show_sort_weight":93,"slug":94},"Story & Novel",90,"story-novel",{"id":45,"doc_module":4,"doc_module_name":44,"category_name":96,"show_sort_weight":97,"slug":98},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":44,"category_name":100,"show_sort_weight":101,"slug":102},"Exam",70,"exam",{"id":104,"doc_module":4,"doc_module_name":44,"category_name":105,"show_sort_weight":106,"slug":107},5,"Comic",60,"comic",{"id":109,"doc_module":4,"doc_module_name":44,"category_name":110,"show_sort_weight":111,"slug":112},6,"Technology",50,"technology",{"id":11,"doc_module":4,"doc_module_name":44,"category_name":12,"show_sort_weight":114,"slug":115},40,"healthcare",{"id":117,"doc_module":4,"doc_module_name":44,"category_name":118,"show_sort_weight":119,"slug":120},8,"Research & Report",30,"research-report",{"id":122,"doc_module":4,"doc_module_name":44,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":44,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":44,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":44,"category_name":135,"show_sort_weight":104,"slug":136},19,"General","general"]