{"id":536,"date":"2026-03-03T11:32:41","date_gmt":"2026-03-03T11:32:41","guid":{"rendered":"https:\/\/www.saiep-paris.fr\/?p=536"},"modified":"2026-10-30T22:37:35","modified_gmt":"2026-10-30T22:37:35","slug":"meet-great-sage-the-super-research-partner-a-full-upgrade-for-the-scientist-centred-xinghe-qizhi-platform","status":"publish","type":"post","link":"https:\/\/www.saiep-paris.fr\/en\/meet-great-sage-the-super-research-partner-a-full-upgrade-for-the-scientist-centred-xinghe-qizhi-platform.html","title":{"rendered":"Here Comes \u201cDasheng\u201d, the Super Research Partner! The Scientist-Centered Xinghe Qizhi Platform Undergoes a Comprehensive Upgrade"},"content":{"rendered":"<p><strong>This afternoon (March 1), the Shanghai Academy of AI for Science (hereinafter \u201cSAIAS\u201d), together with Fudan University, held the Xinghe Qizhi \u00b7 Super Research Partner launch event at Mosu Space in Xuhui District, Shanghai.<\/strong><\/p>\n<p><strong>The event marks a comprehensive upgrade of the Xinghe Qizhi Science AI Open Platform (hereinafter the \u201cXinghe Qizhi Platform\u201d). Building on its existing full-process AI for Science capabilities, the platform has added a highly agentic system for scientific exploration, with the \u201cDasheng\u201d Super Research Partner as its core vehicle, to drive a transformation of research paradigms.<\/strong><\/p>\n<p><strong>Reposted from Shanghai Science and Technology<\/strong><\/p>\n<p><strong>This afternoon (March 1), the Shanghai Academy of AI for Science (hereinafter \u201cSAIAS\u201d), together with Fudan University, held the Xinghe Qizhi \u00b7 Super Research Partner launch event at Mosu Space in Xuhui District, Shanghai.<\/strong><\/p>\n<p><strong>The event marks a comprehensive upgrade of the Xinghe Qizhi Science AI Open Platform (hereinafter the \u201cXinghe Qizhi Platform\u201d). Building on its existing full-process AI for Science capabilities, the platform has added a highly agentic system for scientific exploration, with the \u201cDasheng\u201d Super Research Partner as its core vehicle, to drive a transformation of research paradigms.<\/strong><\/p>\n<p><strong>Jin Li, President of Fudan University, and Zhang Ying, Deputy Secretary-General of the Shanghai Municipal Government, attended the event and delivered speeches. Also present were Xu Zheng, Executive Vice President of Fudan University; Gong Xingao, Chairman of the Academic Committee of Fudan University and academician of the Chinese Academy of Sciences; Ding Xiaodong, Secretary of the Party Committee and Executive Vice President of the Shanghai Innovation Institute; Pan Yan, Deputy Director of the Shanghai Municipal Commission of Economy and Informatization; Qu Wei, Deputy Director of the Shanghai Municipal Commission of Science and Technology; Zhao Zhen, Deputy Director of the Shanghai Municipal Education Commission; Lei Xiaotu, Deputy Director of the Shanghai Meteorological Service; Chen Yong, Deputy Mayor of the Xuhui District Government; Li Xin, Vice President of Shanghai State-owned Capital Investment Co., Ltd.; Liu Shanquan, Vice President of Shanghai Yidian (Group) Co., Ltd.; Zhou Lei, Vice President of Fudan University; Wu Fan, Vice President of the Shanghai Medical College of Fudan University; Zhu Tongyu, Vice President of the Shanghai Medical College of Fudan University; and other heads of relevant units and representatives of enterprises.<\/strong><\/p>\n<p>\u00a0<\/p>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<p><strong>01<\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<p><strong>Young Scientists Lead the R&amp;D<\/strong><\/p>\n<p><strong>\u201cDasheng\u2019s\u201d 72 Transformations Capability System<\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p>\u00a0<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>As the AI-driven transformation of scientific research paradigms accelerates, AI has grown from a mere research aid into a core participant in scientific discovery. As a scientist-centered, full-chain AI for Science open platform, the Xinghe Qizhi Platform confronts the new challenges of AI for Science and introduces the \u201cDasheng\u201d Super Research Partner, further filling in key capabilities on top of what it already offered.<\/p>\n<p>\u00a0<\/p>\n<p>The Xinghe Qizhi Platform was officially launched last July. Centered on real research pathways, it has systematically built a full-chain AI for Science infrastructure covering data, models, computing power, experiments, reasoning and collaboration, forming <strong>an integrated \u201cdata\u2013model\u2013computing power\u2013experiment\u2013agent\u2013community\u201d<\/strong> research intelligence closed loop.<\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11121.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>Over the past seven months the platform&#8217;s capabilities have kept evolving, with notable results. SAIAS, together with Fudan University and other universities and research institutions, has achieved systematic breakthroughs in several disciplines on the Xinghe Qizhi Platform, initially forming <strong>a \u201cjoint research and co-creation\u201d AI for Science ecosystem<\/strong>.<\/p>\n<p>\u00a0<\/p>\n<p>Yet as AI capabilities keep leaping forward, the core problems facing AI for Science are also changing. The challenge is no longer whether models are powerful enough, but how, under professional constraints, to enable agents to complete an entire research workflow with a high degree of autonomy.<\/p>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11122.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>On this basis, the Xinghe Qizhi Platform has been comprehensively upgraded with the launch of the \u201cDasheng\u201d Super Research Partner.<strong><span class=\"wx_text_underline\">With this highly agentic system, \u201cDasheng\u201d, scientists can propose research directions in natural language; Dasheng makes centralized decisions and decomposes tasks, dispatches multiple agents to execute them collaboratively, and deposits structured memory throughout the process<\/span><\/strong><span class=\"wx_text_underline\">; through a two-layer communication and branch governance mechanism, vague instructions are isolated while the system&#8217;s organization stays clear, preventing paths from falling out of order in complex tasks.<\/span><\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11123.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p><strong>Built around real research scenarios, \u201cDasheng\u201d possesses system-level professional capabilities that keep evolving.<\/strong>To address the shortcomings of general large models in scientific modality reasoning, <strong>\u201cDasheng\u201d uses the \u201cShenzhen\u201d science multimodal foundation model as its \u201cbrain\u201d<\/strong>, focusing on multi-scientific-modality input and generation, scientific multimodal alignment and the model&#8217;s capacity for self-reflection. The model can directly read structural, sequence and field information and output it as structures or graphs; natural language serves as the orchestration layer, coordinating complex modalities and reasoning workflows.<\/p>\n<p>\u00a0<\/p>\n<p>It is worth noting that <strong>the Shenzhen model reaches the industry&#8217;s current state of the art in textual scientific understanding and reasoning<\/strong> (compared with Gemini 3 Pro, GPT-5, NatureLM and others), <strong>is on par with specialized models in RNA understanding and generation, and at the same time possesses reasoning capabilities across both text and scientific modalities.<\/strong><\/p>\n<p>\u00a0<\/p>\n<p>One of the core difficulties scientists face today in using AI for research is the lack of \u201cusable capabilities\u201d that can span long workflows, multiple stages and multiple data sources.<strong>To solve this, \u201cDasheng\u201d has a Skills system for many research scenarios that distills key capabilities into executable sub-workflows; it can break down steps, call on computing power and model resources, and continuously iterate on paths to form closed-loop execution.<\/strong><\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11124.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>Existing AI tools face memory challenges in multi-agent collaboration and are prone to error propagation, information overwriting and rework. Research, by contrast, requires collective memory: protecting accurate consensus memory, retaining non-consensus and failed memory, and supporting memory fusion and optimization.<strong>Drawing on Git branch management in software development, \u201cDasheng\u201d builds a multi-branch collective memory architecture that can support long-term collaboration among multiple agents.<\/strong>In this architecture, established theories form the trunk while innovative paths evolve in parallel on independent branches, achieving isolation and full preservation of multiple paths; failed branches retain lessons, successful results are merged into the trunk through controlled merges, collective understanding keeps upgrading, and data and experience can be reused across scenarios.<\/p>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11125.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>As AI enters research scenarios in depth, its capabilities must operate within the rules.<strong>\u201cDasheng\u201d uses safe and trustworthy research agents as a \u201cgolden headband\u201d that guards the boundaries of research security<\/strong>. A systematic security architecture ensures that \u201cdata does not leave and tasks can come in\u201d: Skills handle content security and protect research intellectual property, the Sandbox handles execution-environment security and MCP handles service security, providing protection level by level. Researchers can complete model validation, workflow testing and run research agents in a natively integrated large-scale Sandbox environment, with the system automatically managing permissions and resource lifecycles for low-cost trial and error and rapid validation; mature results then migrate smoothly to the formal computing environment.<\/p>\n<p>\u00a0<\/p>\n<p>On collaborative security, the platform is also exploring <strong>\u201cSafe Agent on Chain\u201d<\/strong> collaboration mechanisms that bring task decomposition, execution traces and result generation into rights-confirmable, verifiable chained records. Combining privacy computing with evidence storage and traceability encourages researchers to dare and be willing to share research data, accelerating the flow of high-value research data.<\/p>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11126.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>In addition, based on autonomous agent exploration and integrated hardware-software orchestration, \u201cDasheng\u201d proposes a whole-domain scientific closed loop that goes \u201cup to the sky and into the body\u201d.<strong>Through the self-driven \u201cLife Lab\u201d and \u201cSky Computing Lab\u201d, it achieves all-weather, cross-scale physical empirical validation that integrates the macro and the micro.<\/strong>The so-called <strong>\u201creaching the sky\u201d<\/strong> means pushing computing capabilities to the frontier of space observation. The Sky Computing Lab works on end-to-end large models, spaceborne intelligent computers, new payloads and inter-satellite networking to improve the real-time use of space data and the effectiveness of weather forecasting. The so-called <strong>\u201centering the body\u201d<\/strong> means going inside living organisms to discern the laws of human representations at the macro and micro scales. In the Life Lab, the system runs on a \u201ccompute\u2013validate\u2013update\u201d mechanism, linking software to hardware in a closed loop through reinforcement learning, so that the model becomes more intelligent the more it learns.<\/p>\n<p>\u00a0<\/p>\n<p>On site, several young core researchers demonstrated in real time to the audience \u201cDasheng\u2019s\u201d \u201c72 Transformations\u201d system-level professional capabilities.<strong>Several of these capabilities are already in practical use and have even reached the industry&#8217;s frontier, and are expected to deliver greater value in the future.<\/strong><\/p>\n<\/section>\n<\/section>\n<section>\n<p>\u00a0<\/p>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<p><strong>02<\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<p><strong>Building on the Xinghe Qizhi Platform<\/strong><\/p>\n<p><strong>Enriching the AI for Science Ecosystem<\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p>\u00a0<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p>\u00a0<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>At the launch event, the innovation track of the 4th World AI for Science Competition \u2014 <strong>the AI4S Agent CNS Challenge \u2014 was released on the same occasion today and registration opened.<\/strong>Hosted by SAIAS and Fudan University, the competition focuses on autonomous research agents: competing agents must, with zero human intervention, independently complete literature comprehension, code reconstruction, hypothesis generation, experimental validation and result iteration, and on that basis challenge and surpass the current state of the art (SOTA) of results published in top journals such as CNS and of research of the same level.<\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11127.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>The competition will run in three stages \u2014 preliminary, semi-final and final \u2014 with four challenge tasks: genome-scale virtual drug screening optimization, a closed loop of targeted molecule design and retrosynthesis planning, a protein conformational ensemble generation challenge, and automatic improvement of neural operators.<\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11128.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<p><strong>To accelerate scientific and technological innovation and the creation of industrial value, SAIAS has taken the lead in launching an AI for Science \u201cAcceleration Camp\u201d.<\/strong>As an important coordination mechanism within the Xinghe Qizhi ecosystem, the camp aims to build an organized research model that deeply integrates universities, research institutions and enterprises, emphasizing joint efforts under \u201cthe same space, the same time and the same goal\u201d and enabling scientists, AI experts and engineers to jointly define problems, set technical paths and iterate models, thereby accelerating scientific discovery. In its first cohort, 86 cross-disciplinary teams applied and 25 were selected, working with Fudan University and other institutions, covering fields such as environmental, earth, materials, life and legal sciences.<\/p>\n<p><img referrerpolicy=\"no-referrer\" decoding=\"async\" src=\"https:\/\/www.saiep-paris.fr\/wp-content\/uploads\/2026\/11\/11129.jpg\" alt=\"\" \/><\/p>\n<\/section>\n<\/section>\n<section>\n<section>\n<p>In addition, to advance the commercialization of AI for Science results, SAIAS signed strategic cooperation agreements at the event with several companies and institutions, including Shanghai State-owned Capital Investment Co., Ltd. and Shanghai Yidian (Group) Co., Ltd., to further strengthen linkages in key areas such as capital coordination, innovation commercialization, data supply, frontier models and the open community, and to build a complete ecosystem.<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section><\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section>\n<section>\n<section>\n<section><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<section>\n<section>\n<section data-mpa-action-id=\"mm7vvyw81gmp\" data-pm-slice=\"0 0 []\"><\/p>\n<section data-mpa-template=\"t\">\n<section data-mpa-template=\"t\">\n<hr \/>\n<p>Source: Fudan University<\/p>\n<p>Editor: Shi<\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<p>\u00a0<\/p>\n<p>\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"<p>This afternoon (March 1), the Shanghai Academy of AI for Science (hereinafter \u201cSAIAS\u201d), together with Fudan University, held the Xinghe Qizhi \u00b7 Super Research Partner launch event at Mosu Space in Xuhui District, Shanghai. The event marks a comprehensive upgrade of the Xinghe Qizhi Science AI Open Platform (hereinafter the \u201cXinghe Qizhi Platform\u201d). Building on its existing full-process AI for Science capabilities, the platform has added a highly agentic system for scientific exploration, with the \u201cDasheng\u201d Super Research Partner as its core vehicle, to drive a transformation of research paradigms. [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[35],"tags":[],"class_list":["post-536","post","type-post","status-publish","format-standard","hentry","category-news-and-trends"],"_links":{"self":[{"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/posts\/536","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/comments?post=536"}],"version-history":[{"count":3,"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/posts\/536\/revisions"}],"predecessor-version":[{"id":2523,"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/posts\/536\/revisions\/2523"}],"wp:attachment":[{"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/media?parent=536"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/categories?post=536"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.saiep-paris.fr\/en\/wp-json\/wp\/v2\/tags?post=536"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}