We’re excited to share <strong>S1-Omni</strong>, a unified multimodal reasoning model for scientific understanding, prediction, and generation.</p>\n<p>S1-Omni brings heterogeneous scientific inputs—including text, material CIFs, molecular SMILES, protein sequences, spectra, and scientific images—into a shared reasoning framework, while using specialized decoders to produce verifiable, domain-native outputs.</p>\n<p>Highlights:</p>\n<ul>\n<li>One unified model spanning 200+ scientific tasks across nine disciplines</li>\n<li>Evaluated on more than 60 scientific benchmarks</li>\n<li>Supports property prediction, spectrum-to-structure reconstruction, protein analysis and structure prediction, and scientific image generation/editing</li>\n<li>Open model weights, inference code, and a 10K-sample subset of the S1-Omni-Corpus</li>\n</ul>\n<p>🔗 <a href=\"https://scienceone-ai.github.io/S1-Omni/\" rel=\"nofollow\">Project page</a><br>🤗 <a href=\"https://huggingface.co/ScienceOne-AI/S1-Omni\">Model weights</a><br>🤗 <a href=\"https://huggingface.co/datasets/ScienceOne-AI/S1-Omni-Corpus-10K\">Dataset</a><br>💻 <a href=\"https://github.com/ScienceOne-AI/S1-Omni\" rel=\"nofollow\">Code</a></p>\n<p>We welcome feedback, questions, and discussions from the AI4Science community!</p>\n","updatedAt":"2026-07-20T06:35:25.374Z","author":{"_id":"641c563fdad248407395bf78","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/641c563fdad248407395bf78/63popJJlkLtqpDmk7xhh4.png","fullname":"Bing","name":"Jeol","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":3,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7447562217712402},"editors":["Jeol"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/641c563fdad248407395bf78/63popJJlkLtqpDmk7xhh4.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.15686","authors":[{"_id":"6a5d8ebc6a69ce099f4d6dc6","name":"Jiahao Zhao","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dc7","name":"Junyi Liu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dc8","name":"Lifeng Xu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dc9","name":"Nan Xu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dca","name":"Qingli Wang","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dcb","name":"Qingxiao Li","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dcc","name":"Tianle Chen","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dcd","name":"Xiaoyu Wu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dce","name":"Yawen Zheng","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dcf","name":"Zikai Wang","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd0","name":"Guanming Liu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd1","name":"Hequn Zhou","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd2","name":"Jingyi Wang","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd3","name":"Jingyuan Shu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd4","name":"Keqi Wang","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd5","name":"Li He","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd6","name":"Songyang Diao","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd7","name":"Wenhui Xu","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd8","name":"Xinyu Ren","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dd9","name":"Yaqin Fan","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6dda","name":"Yujin Zhou","hidden":false},{"_id":"6a5d8ebc6a69ce099f4d6ddb","name":"Zhanao Yao","hidden":false}],"publishedAt":"2026-07-17T00:00:00.000Z","submittedOnDailyAt":"2026-07-20T00:00:00.000Z","title":"S1-Omni: A Unified Multimodal Reasoning Model for Scientific Understanding, Prediction, and Generation","submittedOnDailyBy":{"_id":"641c563fdad248407395bf78","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/641c563fdad248407395bf78/63popJJlkLtqpDmk7xhh4.png","isPro":false,"fullname":"Bing","user":"Jeol","type":"user","name":"Jeol"},"summary":"We present S1-Omni, a unified multimodal reasoning model for scientific understanding, prediction, and generation. AI for Science (AI4S) has advanced significantly through domain-specific models, tool-augmented LLMs, and scientific language models. However, model capabilities remain highly fragmented, limiting the joint modeling of heterogeneous data, scientific laws, and expert knowledge. S1-Omni addresses this gap by consolidating these capabilities into a single, coherent scientific reasoning model. The architecture of S1-Omni is built upon three core components: unified representation of scientific data, natural-world knowledge alignment, and decoding for domain-specific tasks. First, S1-Omni maps natural-language instructions and scientific objects, including CIF, SMILES, protein sequences, spectra, and scientific images, into a shared representation space. Second, it incorporates scientific laws and expert knowledge into data construction and training, enabling the model to reason from scientific evidence. Third, it performs task-specific decoding to support a broad range of applications, including property prediction, spectrum-to-molecular generation, protein site and structure prediction, and scientific image generation and editing. S1-Omni is trained on S1-Omni-Corpus, which covers 200 scientific tasks and contains millions of reasoning samples, and is evaluated on over 60 scientific benchmarks. It outperforms GPT-5.5 and Gemini-3.1-Pro on most benchmarks and matches or surpasses domain-specific models on several benchmarks. Overall, S1-Omni provides a practical path toward unified scientific modeling.","upvotes":4,"discussionId":"6a5d8ebc6a69ce099f4d6ddc","projectPage":"https://scienceone-ai.github.io/S1-Omni/","githubRepo":"https://github.com/ScienceOne-AI/S1-Omni","githubRepoAddedBy":"user","githubStars":7},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"67c8119abb1ec857e1f17fe6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/67c8119abb1ec857e1f17fe6/BY_airx3vQwd4r3CYStXG.png","isPro":false,"fullname":"ScienceOne-AI","user":"ScienceOne-AI","type":"user"},{"_id":"641c563fdad248407395bf78","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/641c563fdad248407395bf78/63popJJlkLtqpDmk7xhh4.png","isPro":false,"fullname":"Bing","user":"Jeol","type":"user"},{"_id":"69ccfba7458769bb4a97a92e","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/jHjL4EW7CfJEhsxOP0fqv.jpeg","isPro":false,"fullname":"Zhou Ziyi","user":"sebastianhil652","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.15686.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2607.15686 in a Space README.md to link it from this page.