单词学习：![[dbeb8605-9b6d-4d6e-abb2-a4b362fd3a36.mp3]]

新闻播报：![[fb238b5b-1159-499c-89bd-60986a11bcff.mp3]]

GPT-5独立解决数学难题，其归因透明化论文引学界激辩

简报：

• 瑞士数学家约翰内斯·施密特宣布，GPT-5在无人类干预下，首次独立解决了一个长期未解的数学难题。
• 为实现完全可追溯，该研究论文对每一部分的贡献者（包括GPT-5、Gemini3Pro、Claude等AI模型及人类）进行了精确标注，并附有原始对话链接。
• 这种极致的透明化归因实验因其过程繁琐，在学界引发了关于其可能成为“学术官僚主义”并阻碍创新的争论。
• 该事件也引发了更深层的讨论：在人类提供初始提示词和筛选结果的情况下，如何界定AI的“独立贡献”，以及未来AI时代学术出版的标准。

相关链接：

• GPT-5 首次独立攻克数学难题，透明化归因实验引发学界争论

GPT-5 Independently Solves Mathematical Problem, Attribution Transparency Paper Sparks Academic Debate 🤯

Brief:

• Swiss mathematician Johannes Schmidt announced that GPT-5 has, for the first time, independently solved a long-standing mathematical problem without human intervention.
• To achieve complete traceability, the research paper precisely attributes contributions for each section (including AI models like GPT-5, Gemini3Pro, Claude, and humans), accompanied by links to original conversations. 🔗
• This experiment in extreme transparent attribution has sparked debate within academia regarding its cumbersome process, with some arguing it could become "academic bureaucracy" and hinder innovation.
• The event has also prompted deeper discussions: how to define "independent contribution" of AI when humans provide initial prompts and filter results, and what academic publishing standards will look like in the future AI era. 🤔

Related Links:

• GPT-5 Independently Solves Mathematical Problem for the First Time, Transparent Attribution Experiment Sparks Academic Debate

文心5.0 Preview创国产模型新高，登顶LMArena榜单

简报：

• 百度新一代模型文心5.0 Preview（ERNIE-5.0-Preview-1203）在全球大模型竞技场LMArena的文本榜单中以1451分登顶，成为排名最高的国产大模型，位列全球并列第二。
• 测评显示，该模型在创意写作和理解高难度复杂指令方面表现突出，生成的文本内容在逻辑一致性上超过了多款国内外主流模型。
• 文心5.0是一款参数量达2.4万亿的原生全模态大模型，采用原生全模态统一建模技术，实现了文本、图像、音频、视频等多种信息的原生融合。

相关链接：

• 国产大模型首登顶！文心5.0 Preview在LMArena全球竞技场拿下中国最高分

ERNIE 5.0 Preview Sets New Record for Domestic Models, Tops LMArena Rankings

Brief:

• Baidu's new-generation model, ERNIE 5.0 Preview (ERNIE-5.0-Preview-1203), has topped the text list in the global large model arena LMArena with 1451 points, becoming the highest-ranked domestic large model and globally tied for second place. 🚀
• Evaluations show that the model excels in creative writing and understanding highly complex instructions. The generated text content surpasses several mainstream domestic and international models in terms of logical consistency. ✨
• ERNIE 5.0 is a native full-modal large model with 2.4 trillion parameters, utilizing native full-modal unified modeling technology to achieve the native fusion of various information types, including text, images, audio, and video. 🧠

Related Links:

• Chinese Large Model Tops for the First Time! ERNIE 5.0 Preview Achieves China's Highest Score in LMArena Global Arena

智谱AI开源新模型GLM-4.7，编程能力创SOTA纪录

简报：

• 智谱AI于2025年12月23日发布新一代旗舰模型GLM-4.7，并宣布上线即开源。
• 该模型主打编程能力，在SWE-Bench基准测试中得分达到73.8%，刷新了开源模型在该项测试中的最高纪录。
• 技术报告显示，GLM-4.7在AIME 25和人类最后考试（HLE）等基准测试中的分数超过了GPT-5.1。
• 新模型在技术上引入了“保留式思考”和“轮级思考”模式，以提升在长程、复杂任务中的表现。
• GLM-4.7现已通过官方网站的Chatbot和API提供服务。

相关链接：

• 智谱发布新一代旗舰AI模型GLM-4.7：编程能力达SOTA，上线即开源

Zhipu AI Open-Sources New Model GLM-4.7, Achieves SOTA Record in Programming Capability

Brief:

• Zhipu AI released its new generation flagship model, GLM-4.7, on December 23, 2025, announcing its immediate open-sourcing upon launch. 🚀
• The model primarily focuses on programming capabilities, achieving a score of 73.8% in the SWE-Bench benchmark test, setting a new record for open-source models in this evaluation.
• According to technical reports, GLM-4.7's scores in benchmarks such as AIME 25 and Human Last Exam (HLE) surpassed those of GPT-5.1. 🤯
• The new model technically introduces "Retentive Thinking" and "Turn-based Thinking" modes to enhance performance in long-term, complex tasks.
• GLM-4.7 is now available via the official website's Chatbot and API. ✨

Related Links:

• Zhipu Launches New Generation Flagship AI Model GLM-4.7: Programming Capability Reaches SOTA, Open-Source Upon Launch

优必选携手字节跳动，共推大模型在工业机器人领域落地

简报： 优必选旗下智慧物流子公司优奇与字节跳动旗下火山引擎正式签署合作协议，双方将深度整合各自在机器人本体与云端AI技术上的优势，并围绕豆包大模型展开合作，共同加速大模型在工业与物流领域的规模化落地。

相关链接：

• 优必选子公司优奇联手火山引擎，豆包大模型赋能具身智能新赛道

UBTECH Partners with ByteDance to Advance Large Models in Industrial Robotics

Brief: Youqi, a smart logistics subsidiary of UBTECH, has officially signed a cooperation agreement with Volcengine, a ByteDance subsidiary. The two parties will deeply integrate their respective strengths in robotic bodies and cloud AI technology, and collaborate around the Doubao large model to jointly accelerate the large-scale implementation of large models in the industrial and logistics sectors. 🤖🚀

Related Link:

• UBTECH's subsidiary Youqi partners with Volcengine, Doubao large model empowers a new track in embodied AI ✨

苹果智能眼镜新爆料：定位iPhone配件，主打AI体验

简报：

• 据彭博社记者马克·古尔曼于12月22日披露，苹果首款智能眼镜计划于2026年底揭晓，2027年正式发货。
• 初代产品设计将优先考虑轻量化，定位为类似Apple Watch的“iPhone智能配件”。
• 该眼镜的核心卖点将围绕Siri语音助手和AI智能体验，而非视觉重叠的增强现实功能。
• 该项目被苹果CEO蒂姆·库克列为“最高战略优先级”，旨在为从Vision Pro等重型头显向日常穿戴设备过渡铺平道路。

相关链接：

• 苹果初代智能眼镜细节曝光，将成iPhone最强AI配件

New Apple Smart Glasses Leaks: Positioned as iPhone Accessory, Focus on AI Experience 👓

Brief:

• According to Bloomberg reporter Mark Gurman's disclosure on December 22nd, Apple's first smart glasses are planned to be unveiled by the end of 2026, with official shipments starting in 2027.
• The design of the first-generation product will prioritize lightweight construction, positioning it as an "iPhone smart accessory" similar to the Apple Watch.
• The core selling points of the glasses will revolve around the Siri voice assistant and AI intelligent experiences, rather than augmented reality features with visual overlays. 🤖
• This project has been designated a "top strategic priority" by Apple CEO Tim Cook, aiming to pave the way for a transition from heavy headsets like the Vision Pro to everyday wearable devices. ✨

Related Links:

• Details of Apple's First Smart Glasses Emerge, Set to Be iPhone's Strongest AI Accessory

腾讯云领跑Valkey社区，国内首发8.0版本

简报：

• 12月23日，腾讯云宣布在国内率先支持并发布开源内存数据库Valkey 8.0版本。
• 腾讯云数据库团队近两年在Redis与Valkey社区的累计贡献量位居全球第一，超过360个PR。
• Valkey是由Linux基金会托管的开源内存数据库，源自Redis社区，致力于构建厂商中立的平台。

相关链接：

• 腾讯云Valkey社区贡献位居全球第一，率先在国内支持8.0版本

Tencent Cloud Leads Valkey Community with First Domestic Release of Version 8.0

Brief:

• On December 23rd, Tencent Cloud announced its pioneering support and release of the open-source in-memory database Valkey version 8.0 in China. 🚀
• The Tencent Cloud Database Team's cumulative contributions to the Redis and Valkey communities have ranked first globally in the past two years, with over 360 Pull Requests. 🥇
• Valkey is an open-source in-memory database hosted by the Linux Foundation, originating from the Redis community, and dedicated to building a vendor-neutral platform. ✨

Related links:

• Tencent Cloud ranks first globally in Valkey community contributions, first to support version 8.0 domestically

Alphabet收购Intersect，为AI算力储备庞大绿色能源

简报：

• 谷歌母公司Alphabet宣布以47.5亿美元（含承担债务）的价格收购清洁能源开发商Intersect，旨在为日益庞大的AI算力提供电力支撑。
• Intersect预计到2028年投入运营的发电规模将达到10.8吉瓦，发电量相当于胡佛水坝的20倍以上，将为Alphabet的数据中心提供动力。
• 收购完成后，Intersect将保持独立运营，并重点探索新兴能源技术，以支持谷歌在美国的基础设施。
• 在此之前，双方已合作开发毗邻清洁能源工厂的工业园区，用于容纳超大规模数据中心。

相关链接：

• Alphabet斥资47.5亿美元收购Intersect，加码AI算力绿色能源

Alphabet Acquires Intersect, Securing Vast Green Energy for AI Computing Power

Briefing:

• Google's parent company, Alphabet, announced its acquisition of clean energy developer Intersect for $4.75 billion (including assumed debt). This move aims to provide power support for the ever-growing AI computing demands. 💡
• Intersect is projected to have 10.8 gigawatts (GW) of operational generation capacity by 2028, which is more than 20 times the power output of the Hoover Dam. This capacity will power Alphabet's data centers. ⚡
• Post-acquisition, Intersect will maintain independent operations and focus on exploring emerging energy technologies to support Google's infrastructure in the United States.
• Prior to this acquisition, the two companies had already collaborated on developing industrial parks adjacent to clean energy plants, designed to house hyperscale data centers. 🌳

Related Links:

• Alphabet Acquires Intersect for $4.75 Billion, Boosting Green Energy for AI Computing Power