Investment Rating - The report does not explicitly state an investment rating for the industry or company Core Insights - DeepSeek-R1 introduces a new paradigm of strong reasoning under reinforcement learning (RL), showcasing significant advancements in reasoning capabilities and long-text processing [4][7] - The model demonstrates exceptional performance in complex tasks, marking a milestone in the open-source community's competition with closed-source models like OpenAI's o1 series [7] - The report highlights the potential of RL-driven models to enhance reasoning abilities without relying on human-annotated supervised fine-tuning [21][56] Summary by Sections Technical Comparison - The report discusses the comparison between STaR-based methods and RL-based methods, emphasizing the advantages of RL in reasoning tasks [3] - It details the innovative RL algorithms used, such as GRPO, which optimize training efficiency and reduce computational costs [49][50] DeepSeek-R1 Analysis - DeepSeek-R1 Zero is built entirely on RL without supervised fine-tuning, showcasing its ability to develop reasoning capabilities autonomously [13][21] - The model's performance metrics indicate strong results in various benchmarks, including AIME 2024 and MATH-500, where it achieved 79.8% and 97.3% respectively, comparable to OpenAI's models [7][15] Insights and Takeaways - The report emphasizes the importance of a robust base model, DeepSeek-V3, which was trained on 671 billion parameters and 14.8 trillion high-quality tokens, enabling significant reasoning capabilities [45][56] - The use of rule-based rewards in training helps avoid reward hacking issues, allowing for automated verification and annotation of reasoning tasks [17][22] Future Directions - The report discusses the potential for further advancements in RL-driven models, suggesting that future training will increasingly focus on RL while still incorporating some supervised fine-tuning [56] - It highlights the need for models to maintain high reasoning performance while ensuring safety and usability in diverse applications [59] Economic and Social Benefits - The exploration of low-cost, high-quality language models is expected to reshape industry dynamics, leading to increased competition and innovation [59] - The report notes that the capital market's volatility is a short-term phenomenon driven by rapid advancements in AI technology, which will lead to a long-term arms race in computational resources [59]

中国小微经营者调查 2024 年四季度报告 暨 2025 年一季度中国小微经营者信心指数报告 Online Survey of Micro-and-small Enterprises (OSOME): Quarterly Report (2024Q4) and Confidence Index (2025Q1) 北京大学企业大数据研究中心 Center for Enterprise Research, Peking University 北京大学中国社会科学调查中心 Institute of Social Science Survey, Peking University 蚂蚁集团研究院 Ant Group Research Institute 网商银行 MYbank 2025 年 2 月 February, 2025 调查报告参与者 张晓波、孔涛、王冉冉、承子珺、陈秋惠、孙秀丽、杨笑寒、向 勖、权盈月、马文利、刘硕 李振华、王芳、谢专 技术支持 数字经济开放研究平台、蚂蚁集团客户体验及权益保障部、网商银行 北京大学小微调研链接： https://cer.gsm.pku.edu.cn/survey/OSOME ...

DeepSeek-R1 \ Kimi 1.5 及 类强推理模型开发解读 陈博远 北京大学2022级"通班" 主要研究方向：大语言模型对齐与可扩展监督 https://cby-pku.github.io/ https://pair-lab.com/ 北大对齐小组 Outline 2 ➢ 技术对比探讨 ➢ DeepSeek-R1 开创RL加持下强推理慢思考范式新边界 ➢ DeepSeek-R1 Zero 及 R1 技术剖析 ➢ Pipeline 总览 \ DeepSeek-V3 Base \ DeepSeek-R1 Zero 及 R1 细节分析 ➢ RL 算法的创新：GRPO及其技术细节 ➢ DeepSeek-R1 背后的Insights & Takeaways：RL加持下的长度泛化 \ 推理范式的涌现 ➢ DeepSeek-R1 社会及经济效益 ➢ STaR-based Methods vs. RL-based Methods 强推理路径对比 (DS-R1 \ Kimi-1.5 \ o-series) ➢ 蒸馏 vs. 强化学习驱动：国内外现有各家技术路线对比分析及Takeaways ➢ PRM & MCTS ...

中国小微经营者调查 2024 年四季度报告 暨 2025 年一季度中国小微经营者信心指数报告 Online Survey of Micro-and-small Enterprises (OSOME): Quarterly Report (2024Q4) and Confidence Index (2025Q1) 北京大学企业大数据研究中心 Center for Enterprise Research, Peking University 北京大学中国社会科学调查中心 Institute of Social Science Survey, Peking University 蚂蚁集团研究院 Ant Group Research Institute 网商银行 MYbank 2025 年 2 月 February, 2025 调查报告参与者 张晓波、孔涛、王冉冉、承子珺、陈秋惠、孙秀丽、杨笑寒、向 勖、权盈月、马文利、刘硕 李振华、王芳、谢专 技术支持 数字经济开放研究平台、蚂蚁集团客户体验及权益保障部、网商银行 北京大学小微调研链接： https://cer.gsm.pku.edu.cn/survey/OSOME ...

北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 北大汇丰智库 摘要 2024 年第三季度，在电子产品等全球需求保持活跃的背景下，东南亚地区经 济增长表现强劲。越南、泰国和新加坡增长超预期，越南的工业生产、消费和出口 强劲增长，新加坡经济受益于半导体需求上升，泰国依靠出口复苏和旅游业实现快 速增长。马来西亚和印尼增速放缓，但仍具韧性，菲律宾则受自然灾害和净出口下 降拖累。CPI 总体下降，主要受国际粮价和能源价格回落影响。 东南亚六国的进出口贸易显著增长，对美出口表现突出。短期内，东南亚可 能继续从中美贸易格局中受益，但中长期来看，贸易保护主义和中美紧张局势可能 带来外部冲击。 中国与东盟关系日益密切，双方已结束自贸区 3.0 版谈判，且中印尼、中新和 中马的 ...

Investment Rating - The report does not explicitly provide an investment rating for the industry Core Insights - The report emphasizes the significant socio-economic challenges posed by the green transition in coal-dependent regions of China, particularly in Shanxi Province, which holds over 20% of the country's coal reserves and employs 37% of the national coal workforce [11][12] - It highlights the need for a coordinated national mechanism to support regions like Shanxi in their transition away from coal, ensuring that the impacts of the green transition are addressed equitably across different provinces [13][14] Summary by Sections Executive Summary - The report outlines China's commitment to peak carbon emissions by 2030 and achieve carbon neutrality by 2060, indicating a shift away from coal dependency [11] - It predicts a direct loss of 320,000 to 330,000 jobs in Shanxi's coal sector between 2025 and 2030, with potential reductions of 96%-98% in coal labor by 2060 under various scenarios [12] Chapter 1: Research Background - The chapter discusses the uneven progress of green transition across different regions in China, with significant disparities in resource endowments and socio-economic development [18][20] Chapter 2: Analysis of Key Transition Regions - This chapter assesses the regional distribution of coal resources and employment, identifying Shanxi, Inner Mongolia, and Shaanxi as critical areas facing severe employment challenges due to their reliance on coal [22][30] - It notes that Shanxi's coal employment numbers are significantly higher than other provinces, with 800,000 workers in the coal sector, which is four times that of Henan [30] Chapter 3: Economic and Employment Impacts of Energy Transition in Shanxi - The chapter details the contribution of coal to Shanxi's economic growth, noting a strong correlation between GDP growth and coal consumption [48] - It highlights that the coal sector's revenue contribution to the industrial sector has increased, reaching 42% by 2022 [50] Chapter 4: Transition Experiences from Domestic Regions - This section presents successful transition experiences from other resource-dependent cities, emphasizing the importance of diversifying the economy and stabilizing the labor market during structural changes [12][14] Chapter 5: Policy Recommendations - The report suggests establishing a regional support mechanism at the national level to mitigate the differentiated impacts of the green transition and encourages pilot explorations in key transition areas [13][14] - It advocates for policies that promote economic diversification and support for affected workers, including job placement services and social security enhancements [13][14]

Investment Rating - The report does not explicitly provide an investment rating for the industry Core Insights - The report emphasizes the importance of digital ecology in enhancing new productive forces, which is crucial for high-quality development in China. It highlights the relationship between digital ecology and new productive forces, exploring topics such as data assetization, pricing mechanisms in the computing power market, and the role of data intelligence in promoting low-carbon development in the construction sector [4][14][32]. Summary by Sections Digital Ecology Index - The 2024 Digital Ecology Index indicates a stable four-tier development pattern among provincial regions, with significant competition among cities. The eastern region leads in digital ecology, while the central region is emerging, and the western and northeastern regions lag behind. Major city clusters like Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta form self-circulating digital ecological hubs, while Chengdu-Chongqing and Central Triangle lack leading cities [4][8]. International Digital Ecology Index - The 2024 International Digital Ecology Index measures 159 countries, showing that China and the United States continue to be the two poles in global digital development. Although China's total index is slightly lower than that of the U.S., it surpasses in digital capability metrics, ranking first globally. European countries lead overall, with Asia following closely [4][8]. Data Assetization - The report discusses the trend of data assetization, which is becoming a significant topic. It notes that over a hundred companies are beginning to recognize data as an asset, which is essential for creating new economic growth points and reducing institutional transaction costs. The development of artificial intelligence and other technologies is crucial for nurturing a healthy computing power market [14][30]. Computing Power Market - The rapid growth of computing power, especially intelligent computing power, is highlighted as a key driver for the digital economy. The report stresses the need for a standardized, fair, and transparent pricing mechanism for computing power to facilitate its healthy development and integration into various industries [8][31]. Low-Carbon Development - The report outlines the significance of low-carbon transformation in the construction industry, which is vital for achieving national carbon neutrality goals. It emphasizes that data elements can accelerate this transformation and provide new opportunities for sustainable development [9][32]. Digital Governance and Policy - The establishment of data bureaus across various regions is aimed at promoting the overall construction of local digital ecologies and breaking through development bottlenecks. The report suggests that a well-functioning digital ecology relies on the collaboration of various stakeholders and the establishment of a cooperative and win-win mechanism [28][33]. Future Directions - The report indicates that enhancing digital capabilities, creating strategic cities, and accelerating digital integration are crucial for the further development of regional digital ecologies. It also points out that the digital economy's growth presents new challenges regarding energy consumption and carbon emissions [4][14][32].

中国小微经营者调查 2024 年一季度报告 暨 2024 年二季度中国小微经营者信心指数报告 Online Survey of Micro-and-small Enterprises (OSOME): Quarterly Report (2024Q1) and Confidence Index (2024Q2) 北京大学企业大数据研究中心 Center for Enterprise Research, Peking University 北京大学中国社会科学调查中心 Institute of Social Science Survey, Peking University 蚂蚁集团研究院 Ant Group Research Institute 网商银行 MYbank 2024 年 5 月 ...