https://github.com/sjtu-sai-agents/X-Master?tab=readme-ov-file

\n

?
I kindly note: great insight, but without shared code, it’s unverifiable and less useful.

\n","updatedAt":"2025-07-12T19:37:05.127Z","author":{"_id":"637e260672390afecb10bee7","avatarUrl":"/avatars/202701363fed4e5f06821187f0c8ec4b.svg","fullname":"emilio Frittrang","name":"usewri101","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8331852555274963},"editors":["usewri101"],"editorAvatarUrls":["/avatars/202701363fed4e5f06821187f0c8ec4b.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2507.05241","authors":[{"_id":"6870a7c6c8391850d60978ca","name":"Jingyi Chai","hidden":false},{"_id":"6870a7c6c8391850d60978cb","name":"Shuo Tang","hidden":false},{"_id":"6870a7c6c8391850d60978cc","name":"Rui Ye","hidden":false},{"_id":"6870a7c6c8391850d60978cd","name":"Yuwen Du","hidden":false},{"_id":"6870a7c6c8391850d60978ce","name":"Xinyu Zhu","hidden":false},{"_id":"6870a7c6c8391850d60978cf","user":{"_id":"64c892530d3d1b209d198f41","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/ltIBrJI2TTRjljJWwTV48.jpeg","isPro":false,"fullname":"周梦成","user":"SleeeepyZhou","type":"user"},"name":"Mengcheng Zhou","status":"claimed_verified","statusLastChangedAt":"2025-08-20T19:13:39.971Z","hidden":false},{"_id":"6870a7c6c8391850d60978d0","name":"Yanfeng Wang","hidden":false},{"_id":"6870a7c6c8391850d60978d1","name":"Weinan E","hidden":false},{"_id":"6870a7c6c8391850d60978d2","name":"Yuzhi Zhang","hidden":false},{"_id":"6870a7c6c8391850d60978d3","name":"Linfeng Zhang","hidden":false},{"_id":"6870a7c6c8391850d60978d4","name":"Siheng Chen","hidden":false}],"publishedAt":"2025-07-07T17:50:52.000Z","submittedOnDailyAt":"2025-07-11T04:29:26.187Z","title":"SciMaster: Towards General-Purpose Scientific AI Agents, Part I.\n X-Master as Foundation: Can We Lead on Humanity's Last Exam?","submittedOnDailyBy":{"_id":"62d22496c58f969c152bcefd","avatarUrl":"/avatars/76c3b70e312f25e1e610473475553c5c.svg","isPro":false,"fullname":"Tiezhen WANG","user":"xianbao","type":"user"},"summary":"The rapid advancements of AI agents have ignited the long-held ambition of\nleveraging them to accelerate scientific discovery. Achieving this goal\nrequires a deep understanding of the frontiers of human knowledge. As such,\nHumanity's Last Exam (HLE) provides an exceptionally challenging touchstone for\nevaluating scientific AI agents. In this work, we aim to construct the\nfoundational architecture for general-purpose agents and validate the\ncapabilities through leading performance on HLE. To achieve this, we introduce\nX-Master, a tool-augmented reasoning agent designed to emulate human\nresearchers by interacting flexibly with external tools during its reasoning\nprocess. This agent, guided by the conceptualization of code as an interaction\nlanguage, can flexibly leverage built-in Python libraries and our customized\ntools to augment the reasoning. We further scale its capabilities through\nX-Masters, a scattered-and-stacked agentic workflow that systematically\nenhances breadth and depth of reasoning. Our open-source solution, X-Masters,\nsets a new state-of-the-art record on HLE with a score of 32.1%, surpassing\nOpenAI's and Google's Deep Research (26.6% and 26.9%) and becoming the first to\nexceed the 30% threshold. This work allows us to gain a deeper understanding of\ncomplex task-solving and accumulates valuable experience that can inform future\nadvancements, guiding subsequent model training.","upvotes":5,"discussionId":"6870a7c6c8391850d60978d5","githubRepo":"https://github.com/sjtu-sai-agents/X-Master","githubRepoAddedBy":"auto","ai_summary":"X-Master, a tool-augmented reasoning agent using Python libraries and customized tools, achieves state-of-the-art performance on Humanity's Last Exam with a score of 32.1%.","ai_keywords":["tool-augmented reasoning agent","X-Master","X-Masters","scattered-and-stacked agentic workflow","interaction language","Python libraries","conceptualization of code"],"githubStars":307},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"62d22496c58f969c152bcefd","avatarUrl":"/avatars/76c3b70e312f25e1e610473475553c5c.svg","isPro":false,"fullname":"Tiezhen WANG","user":"xianbao","type":"user"},{"_id":"6270324ebecab9e2dcf245de","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6270324ebecab9e2dcf245de/cMbtWSasyNlYc9hvsEEzt.jpeg","isPro":false,"fullname":"Kye Gomez","user":"kye","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"6575702b15b1ca184b0b2700","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6575702b15b1ca184b0b2700/O9cEodqQmG-gyqMiO_edR.jpeg","isPro":false,"fullname":"Zaibin Zhang","user":"MrBean2024","type":"user"},{"_id":"686db5d4af2b856fabbf13aa","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/6BjMv2LVNoqvbX8fQSTPI.png","isPro":false,"fullname":"V bbbb","user":"Bbbbbnnn","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">