Core Viewpoint - The "Yunfan" unmanned vessel developed by Shanghai Jiao Tong University successfully conducted close-range meteorological observations of Typhoon "Butterfly" near Hainan Island, demonstrating its capabilities in extreme marine weather monitoring [3][4]. Group 1: Technology and Innovation - The "Yunfan" unmanned vessel is a result of ten years of technological breakthroughs in wave energy utilization, wind-wave hybrid drive, and multi-source energy management by a research team led by Professors Tian Xinliang and Wang Peng [4]. - The vessel features a surface buoy, underwater traction device, and a sail propulsion system, allowing it to convert wave motion into forward thrust while utilizing solar panels for energy [4][5]. Group 2: Operational Performance - During its mission, the "Yunfan" maintained a control radius of less than 100 meters and successfully tracked the typhoon's path, with a tracking error of less than 50 meters [3]. - The vessel recorded significant meteorological data, including maximum wave heights exceeding 6 meters and wind speeds over 30 knots, effectively capturing the atmospheric changes in the typhoon's core area [3]. Group 3: Future Applications - The "Yunfan" unmanned vessel is capable of autonomous operation at sea for several months without external energy, making it suitable for various applications such as oceanographic monitoring, ecological surveys, and resource development [5]. - It is expected to contribute significantly to China's digital and smart ocean initiatives, enhancing marine observation and early warning capabilities [5].

Core Insights - The total budget for instrument procurement intentions from various universities in China exceeds 3.4 billion yuan, indicating a significant investment in research and development [2][3] Group 1: Procurement Overview - A total of 34 universities and research institutions have announced their procurement plans for instruments, with a budget exceeding 3.4 billion yuan [3] - The procurement period spans from early 2025 to late 2025, with various universities planning to acquire instruments in different time frames [4][5] Group 2: Notable University Procurement Plans - Tongji University plans to procure instruments worth 127 million yuan between March and May 2025 [4] - Tsinghua University has the highest procurement budget at 370 million yuan scheduled from April to June 2025 [4] - The Chinese Academy of Sciences' South China Sea Institute of Oceanology has a substantial procurement plan amounting to 827 million yuan from April to June 2025 [5] Group 3: Additional Insights - The procurement intentions reflect a broader trend of equipment upgrades initiated since 2024, with a focus on enhancing research capabilities across various disciplines [3] - The recent policy updates in Beijing regarding equipment update loan interest subsidies may further stimulate procurement activities among universities [3]

关注公众号，点击公众号主页右上角" ··· "，设置星标 "⭐" ，关注 鑫椤锂电 资讯~ 2025（第三届）中国固态电池 论坛议题 I C C S I N O | 发言嘉宾 | 题目 | | --- | --- | | ICC鑫椤资讯 | 欢迎致辞 | | 总经理 连萍 | | | 国联汽车动力电池研究院有限责任公司 | 全固态电池工程技术难点 | | 研究院院长 杨容 | | | 上海移移科技有限公司 待定 | 硅基负极在固态电池中的应用现状及前景 | | 天津国安盟固利新材料科技有限公司 待定 | 富锂锰基正极材料在固态电池上的应用 | | 颁奖 茶歇 | | | 深圳新宙邦科技股份有限公司固态子公司 董事长 王希敏 | 固态电池新材料开发进展 | | 深圳市星源材质科技股份有限公司 | 星源固态电解质膜解决方案 | | 研发总监 郭辰 博士 | | | 上海恩捷新材料科技有限公司 待定 | 待定 | | 浙江金羽新能源科技有限公司 蕃事长 黄杜斌 | 待定 | | 午餐 | | | tire L L = + = | | | 中行元二两住战益防分所 温兆银 研究员 | 待定 | | --- | --- ...

Core Viewpoint - The article discusses the ongoing reforms in China's higher education system, emphasizing the need for better integration between academia and industry to address the long-standing issue of disconnection between university education and societal needs [2][3][4]. Group 1: Higher Education Reform - China's higher education is entering a new exploratory phase of reform, with the Ministry of Education announcing the establishment of ten new universities, including Greater Bay Area University and Ningbo Oriental Institute of Technology, focusing on research-oriented education [2]. - The reform efforts are characterized by the emergence of new research-oriented universities and the restructuring of established institutions, such as Fudan University and Shanghai Jiao Tong University, to enhance their educational offerings [2][3]. - The primary focus of these reforms includes the establishment of new research-oriented universities, enterprise-run universities, new engineering disciplines, and vocational technical colleges, aiming to address the historical emphasis on theoretical knowledge over practical application [2][3]. Group 2: Market Demand and Educational Needs - The rapid advancement of technologies like AI is reshaping traditional economic and social structures, necessitating a shift in higher education to meet evolving market demands [3][4]. - The Chinese government has outlined a comprehensive education plan for 2024-2035, emphasizing the integration of education, technology, and talent development to support modernization efforts [3]. - Despite producing over one-third of the world's engineering graduates, China faces a structural shortage of innovative and interdisciplinary talent, highlighting the importance of new engineering disciplines and research-oriented universities [4]. Group 3: Challenges and Opportunities - The article points out the critical challenge of integrating education with industry needs, stressing that simply placing students in internships is insufficient for fostering interdisciplinary problem-solving skills [4][5]. - There is a concern that the trend of vocational education transitioning to degree programs may dilute the practical focus of vocational training, raising questions about the effectiveness of such reforms [4][5]. - The emphasis on foundational research is crucial, as it distinguishes human capabilities from artificial intelligence, suggesting that universities should leverage their strengths to enhance educational outcomes [5][6].

Core Viewpoint - The article emphasizes the importance of medical education as a comprehensive process that involves not only knowledge acquisition but also the cultivation of compassion and responsibility in future doctors [1]. Group 1: Medical Education Models - The training models for medical students in China have evolved based on national needs, initially focusing on quickly supplementing healthcare personnel during periods of scarcity [2]. - Various training models exist, including a 5-year undergraduate program, a "5+3" model for obtaining multiple qualifications, an 8-year combined bachelor's and master's program, and a "3+2" model for training specialists [3][4]. - The emergence of new training models is driven by the development needs of hospitals and the integration of medical and engineering fields [4]. Group 2: Pathways for Non-Medical Graduates - Non-medical graduates can still pursue a career in medicine through specific pathways such as re-examination, traditional Chinese medicine apprenticeship, or special programs [6]. - The "4+4" clinical medicine doctoral training model, inspired by the U.S. system, allows non-medical bachelor’s degree holders to enter medical education [6]. Group 3: Training and Development of Doctors - The process of becoming a qualified doctor involves multiple stages of education, including undergraduate education, post-graduate training, and continuing education, all of which are interconnected [6]. - Young resident physicians undergo standardized training before they can perform surgeries independently, with varying timelines for different types of procedures [10][11]. Group 4: The Role of Research in Medicine - Research-oriented doctors play a crucial role in bridging clinical practice and scientific exploration, contributing to the development of new treatments and solutions for medical challenges [12]. - The collaboration between clinical and research-oriented doctors is essential for advancing medical innovation and improving patient care [12]. Group 5: Doctor-Patient Relationship - The cultivation of medical ethics and compassion is vital in medical education, as it directly impacts the doctor-patient relationship [13]. - Establishing a robust ethical framework in medical training is necessary to ensure that future doctors embody the values of care and responsibility [13].

Core Viewpoint - The IDEAL method proposed by a joint team from Shanghai Jiao Tong University and Shanghai AI Lab significantly enhances the performance of large language models (LLMs) across various domains by adjusting the composition of the supervised fine-tuning (SFT) training dataset [3][4]. Group 1: Methodology - The IDEAL method focuses on preparing high-quality training datasets for different domains and modeling the optimization problem to minimize validation loss [5]. - The quantity of training data during the SFT phase is not the key factor; rather, the appropriate distribution of data is crucial to avoid exacerbating the "偏科" phenomenon in models [6][15]. - The research quantifies the impact of data adjustment on the optimal model's performance in the validation set, providing a theoretical foundation for the IDEAL approach [7]. Group 2: Computational Efficiency - The paper employs K-FAC theory to approximate the inverse of the Hessian matrix, which simplifies the computation and allows for scalability to LLM parameter sizes [8]. Group 3: Experimental Results - The IDEAL method was tested on the Llama 3.1 8B model, demonstrating a significant improvement in coding capabilities after just two iterations, regardless of the epoch [10]. - The initial distribution of training data can be further optimized, as IDEAL consistently improved average results across various benchmarks, regardless of the initial distribution [11]. Group 4: Practical Applications - IDEAL addresses the challenge of how to effectively combine high-quality training data from various domains into a unified training set, thus eliminating the need for manual adjustments [14]. - The paper suggests that the optimal value for the hyperparameter m should be around 0.15, as it balances the need for data distribution optimization without being too aggressive [15].

IDEAL团队 投稿 量子位 | 公众号 QbitAI 大幅缓解LLM偏科，只需调整SFT训练集的组成。 本来不擅长coding的Llama 3.1-8B，代码能力明显提升。 上海交大&上海AI Lab联合团队提出创新方法 IDEAL ，可显著提升LLM在多种不同领域上的综合性能。 此外，研究还有一些重要发现，比如： 具体来看—— SFT后LLM部分能力甚至退化 大型语言模型 (LLM) 凭借其强大的理解和逻辑推理能力，在多个领域展现了惊人的能力。除了模型参数量的增大， 高质量的数据是公认的LLM性能提升最关键的影响因素。 当对模型进行监督微调（SFT）时，研究人员发现 LLM在多任务场景下常出现"偏科"现象 ——部分能力突出而部分 能力并未涨进，甚至退化。这种不平衡的现象导致大模型在不同的领域上能力不同，进而影响用户体验。 上海交大和上海AI Lab的研究者迅速将目光聚焦到SFT训练的训练集上，是否可以通过调整训练集的组成来缓解LLM 偏科的情况？直觉上来看，直接将LLM的弱势科目的训练数据增加一倍，就可以让最后的结果发生变化。但是，由于 训练数据之间的耦合关系，研究者通过建模量化每个领域数据对于最终结果的 ...

东星医疗（301290）公告，控股子公司常州东星生物医药有限公司与上海交通大学签订了《技术开发 (委托)合同》，合同总金额共计600万元。合同包括截断(短链)重组Ⅲ型人源化胶原蛋白和全长(长链)重 组Ⅲ型人源化胶原蛋白两个项目，每个项目合同约定的技术开发(委托)总额为300万元。 ...

Core Insights - The "Nezha" team from Shanghai Jiao Tong University and the Second Institute of Oceanography have successfully conducted a joint scientific application test of a sea-air cross-domain vehicle in Qiandao Lake, Zhejiang, validating the excellent performance of the "Nezha" series in multi-medium environments [3][4] - The "Nezha" team has developed two specialized devices: "Nezha-Flying Sub" for comprehensive operations including aerial flight and underwater navigation, and "Nezha-Flying Drift" which integrates aerial and surface platform capabilities for extensive monitoring tasks [3] - The test demonstrated the superior system performance of both devices, with "Nezha-Flying Sub" successfully completing key tasks such as water profile observation and automatic water sampling, while "Nezha-Flying Drift" efficiently performed surface water monitoring and multi-point sampling [3] Industry Developments - This test marks a significant advancement in China's sea-air cross-domain observation and sampling technology, with ongoing collaboration aimed at enhancing device adaptability and reliability in complex marine environments [4] - The focus will be on promoting the industrial application of this technology in areas such as marine risk and disaster early warning, as well as ecological protection, contributing to the construction of a maritime power [4]

Pengfei Liu 投稿 量子位 | 公众号 QbitAI AI能够 独自完成 医疗场景下的诊断任务吗？ 在真实的临床环境中，医生需要 综合分析 大量的患者信息——包括主诉症状、既往病史、体格检查以及各类辅助检查结果，才能 逐步构建出对病情的全面认知。 这一过程不仅要求强大的信息整合能力，更涉及复杂的推理判断。随着大语言模型在复杂推理能力上的不断突破，AI在应对各种 科学挑战的前景也愈发广阔。那么， 在高度依赖专业知识与临床经验的医疗领域 ，AI是否也能胜任"诊断"这一关键任务？ 为系统评估AI在临床诊断任务中的实际表现，来自上海交通大学的SPIRAL Lab与GAIR Lab共同构建了 DiagnosisArena —— 一个用于严格评估AI在专业医学诊断中能力水平的基准测试。 研究团队在DiagnosisArena上对现有多个大语言模型进行测试。 测试结果显示： 即使是o3，在此项高挑战性诊断任务中也只达到了51.12%的准确率 ，而其他开源模型甚至难以取得25%的准确 率。此项结果反映出当前模型在复杂医疗推理任务中仍面临诸多瓶颈。 构造过程：如何打造一个考验医学诊断推理能力的基准？ 研究团队设计了一 ...