{
  "title": "2026-09-04 AI日更 | OpenAI 以 10 亿美元押注 AI 安全防御，Astra 进入法律与游戏工作流",
  "url": "https://miaok.ong/ai-daily/ai-daily-2026-09-04/",
  "date": "2026-09-04T07:00:00+08:00",
  "lastmod": "2026-09-04T07:00:00+08:00",
  "type": "ai-daily",
  "kind": "page",
  "language": "zh",
  "description": "OpenAI 一边推出面向关键基础设施的 Daybreak 计划，强化网络防御与行业落地，一边用 GPT-6 Astra 进入法律检索、财务审阅和游戏原型场景。今天的重点不在模型噱头，而在 AI 正加速嵌入专业流程，工程瓶颈也更清晰地暴露在评测、记忆和可信执行上。",
  "keywords": null,
  "tags": [],
  "categories": [],
  "author": "孔淼",
  "image": "https://miaok.ong/images/avatar.jpg",
  "content": "\u003ch1 id=\"2026-09-04-ai日更--openai-以-10-亿美元押注-ai-安全防御astra-进入法律与游戏工作流\"\u003e\n  2026-09-04 AI日更 | OpenAI 以 10 亿美元押注 AI 安全防御，Astra 进入法律与游戏工作流\n  \u003ca class=\"heading-link\" href=\"#2026-09-04-ai%e6%97%a5%e6%9b%b4--openai-%e4%bb%a5-10-%e4%ba%bf%e7%be%8e%e5%85%83%e6%8a%bc%e6%b3%a8-ai-%e5%ae%89%e5%85%a8%e9%98%b2%e5%be%a1astra-%e8%bf%9b%e5%85%a5%e6%b3%95%e5%be%8b%e4%b8%8e%e6%b8%b8%e6%88%8f%e5%b7%a5%e4%bd%9c%e6%b5%81\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h1\u003e\n\u003cblockquote\u003e\n\u003cp\u003eOpenAI 一边推出面向关键基础设施的 Daybreak 计划，强化网络防御与行业落地，一边用 GPT-6 Astra 进入法律检索、财务审阅和游戏原型场景。今天的重点不在模型噱头，而在 AI 正加速嵌入专业流程，工程瓶颈也更清晰地暴露在评测、记忆和可信执行上。\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch2 id=\"-本期-watch-list-深度导读\"\u003e\n  📖 本期 Watch List 深度导读\n  \u003ca class=\"heading-link\" href=\"#-%e6%9c%ac%e6%9c%9f-watch-list-%e6%b7%b1%e5%ba%a6%e5%af%bc%e8%af%bb\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003cp\u003e今天最值得跟进的，是“AI 正在从通用模型走向可落地系统”这一条主线。Google DeepMind 的 WeatherNext 3 代表基础模型继续向高精度行业预测推进，值得看它如何把天气这种强时空问题做成可部署能力；OpenAI 则把 GPT-6 Astra 用到法律检索和游戏原型里，说明 agent 正在进入专业工作流和生产效率场景。另一组论文更值得工程团队细读：围绕评测觉察、持久记忆、静态 LLM API 的数据层、以及多 agent 证据可信度的研究，集中暴露了当前智能体系统最现实的工程瓶颈。\u003c/p\u003e\n\u003ch2 id=\"-x-平台-ai-热点快讯\"\u003e\n  🌐 X 平台 AI 热点快讯\n  \u003ca class=\"heading-link\" href=\"#-x-%e5%b9%b3%e5%8f%b0-ai-%e7%83%ad%e7%82%b9%e5%bf%ab%e8%ae%af\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003ch3 id=\"话题-1openai-launches-gpt-6-astra-as-most-capable-model-yet\"\u003e\n  话题 1:OpenAI Launches GPT-6 Astra as Most Capable Model Yet\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-1openai-launches-gpt-6-astra-as-most-capable-model-yet\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · News\u003c/li\u003e\n\u003cli\u003e概况:热度时间:5 hours ago,相关帖子数:79000\u003c/li\u003e\n\u003cli\u003e是什么事:X 上热议 OpenAI 预发布名为 Astra 的新模型安全更新，并传出其可能在近日上线、能力显著强于 GPT-5.6 的消息。\u003c/li\u003e\n\u003cli\u003e为什么重要:如果属实，这意味着 OpenAI 可能在模型能力、推理效率和网络安全攻防能力上再次拉开差距，也会影响行业对新一代前沿模型发布时间、命名和发布节奏的判断。\u003c/li\u003e\n\u003cli\u003e讨论概况:讨论焦点主要集中在 Astra 是否已进入最终发布阶段、真实公开名称到底是 GPT-6 还是其他版本、以及泄露信息和官方信息之间有多大可信度；另一部分争议在于现有公开证据更多支持其在网络安全场景突破，而非通用能力全面跃升。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-2lululemon-shares-plunge-15-after-weak-earnings-and-cut-outlook\"\u003e\n  话题 2:Lululemon Shares Plunge 15% After Weak Earnings and Cut Outlook\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-2lululemon-shares-plunge-15-after-weak-earnings-and-cut-outlook\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · News\u003c/li\u003e\n\u003cli\u003e概况:热度时间:3 hours ago,相关帖子数:3800\u003c/li\u003e\n\u003cli\u003e是什么事:Lululemon 发布低于预期的财报并下调业绩展望，股价盘中大跌约 15%。\u003c/li\u003e\n\u003cli\u003e为什么重要:这类消费品牌的业绩与指引变化，常被视为宏观需求、零售数据和市场风险偏好的风向标，也会影响 AI 相关投资对消费科技与零售自动化叙事的估值判断。\u003c/li\u003e\n\u003cli\u003e讨论概况:X 上的讨论主要集中在业绩走弱的原因是需求放缓、竞争加剧还是库存与折扣压力，也有人争论这只是短期波动，还是公司增长逻辑已经明显降温。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-3jd-vance-blames-iran-attacks-for-high-gas-prices-in-white-house-briefing\"\u003e\n  话题 3:JD Vance Blames Iran Attacks for High Gas Prices in White House Briefing\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-3jd-vance-blames-iran-attacks-for-high-gas-prices-in-white-house-briefing\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Other\u003c/li\u003e\n\u003cli\u003e概况:热度时间:1 day ago,相关帖子数:37000\u003c/li\u003e\n\u003cli\u003e是什么事:白宫简报会上，JD Vance 将高油价归因于对伊朗的袭击，引发 X 平台上的广泛关注与转述。\u003c/li\u003e\n\u003cli\u003e为什么重要:这类表态会影响市场对能源价格、地缘政治风险和政策叙事的判断，而这些变量也会间接影响 AI 产业的算力成本、供应链预期和宏观投资环境。\u003c/li\u003e\n\u003cli\u003e讨论概况:X 上的讨论主要集中在这番归因是否成立、是否是在为油价上涨寻找政治解释，以及伊朗局势与能源市场之间的实际关联有多大。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-4john-ternus-takes-over-as-apples-new-ceo-from-tim-cook\"\u003e\n  话题 4:John Ternus Takes Over as Apple\u0026rsquo;s New CEO from Tim Cook\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-4john-ternus-takes-over-as-apples-new-ceo-from-tim-cook\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · News\u003c/li\u003e\n\u003cli\u003e概况:热度时间:2 days ago,相关帖子数:267000\u003c/li\u003e\n\u003cli\u003e是什么事:有消息称，John Ternus 接替 Tim Cook 出任苹果新任 CEO。\u003c/li\u003e\n\u003cli\u003e为什么重要:苹果是全球最重要的科技公司之一，其管理层变化会直接影响 AI 产品路线、芯片策略、设备生态和行业竞争格局。\u003c/li\u003e\n\u003cli\u003e讨论概况:X 上的讨论主要集中在这是否意味着苹果将更激进推进端侧 AI、是否会调整现有的保守产品节奏，以及新 CEO 能否在 AI 时代延续苹果的增长和生态优势。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-5jd-vance-details-trucking-fraud-crackdown-shutting-down-2000-cdl-mills\"\u003e\n  话题 5:JD Vance Details Trucking Fraud Crackdown Shutting Down 2,000 CDL Mills\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-5jd-vance-details-trucking-fraud-crackdown-shutting-down-2000-cdl-mills\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Other\u003c/li\u003e\n\u003cli\u003e概况:热度时间:1 day ago,相关帖子数:46000\u003c/li\u003e\n\u003cli\u003e是什么事:JD Vance公开谈及整治卡车驾驶证（CDL）造假和“CDL工厂”，称相关执法已关闭约2000家违规机构。\u003c/li\u003e\n\u003cli\u003e为什么重要:这类监管行动会影响物流与运输行业的数据可信度、合规成本和人才供给，也会间接影响自动驾驶货运、车队管理等AI应用落地时所依赖的安全与身份验证体系。\u003c/li\u003e\n\u003cli\u003e讨论概况:X上的讨论主要集中在两点：一是支持强力打击欺诈、提升道路安全；二是质疑关闭规模是否过大，以及这会不会进一步加剧卡车司机短缺和行业用工压力。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-6nba-hits-clippers-with-historic-penalties-in-kawhi-leonard-cap-probe\"\u003e\n  话题 6:NBA Hits Clippers with Historic Penalties in Kawhi Leonard Cap Probe\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-6nba-hits-clippers-with-historic-penalties-in-kawhi-leonard-cap-probe\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Sports\u003c/li\u003e\n\u003cli\u003e概况:热度时间:1 day ago,相关帖子数:265000\u003c/li\u003e\n\u003cli\u003e是什么事:NBA因快船队涉嫌在招募和签下科怀·伦纳德过程中规避工资帽，对球队处以创纪录处罚。\u003c/li\u003e\n\u003cli\u003e为什么重要:该事件本身并非人工智能事件，但体现了体育联盟在复杂数据、合同审查和规则执行中对透明度与治理机制的重视，也可能影响体育分析和商业决策类AI应用所依赖的数据环境。\u003c/li\u003e\n\u003cli\u003e讨论概况:X上的讨论主要集中在处罚是否与违规程度相称、快船管理层应承担多大责任、联盟调查和执法是否一致，以及这一判罚是否会成为今后处理工资帽规避案件的先例。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-7liverpool-names-25-man-champions-league-squad-omits-chiesa-and-endo\"\u003e\n  话题 7:Liverpool Names 25-Man Champions League Squad, Omits Chiesa and Endo\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-7liverpool-names-25-man-champions-league-squad-omits-chiesa-and-endo\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Other\u003c/li\u003e\n\u003cli\u003e概况:热度时间:3 hours ago,相关帖子数:4600\u003c/li\u003e\n\u003cli\u003e是什么事:利物浦公布了25人欧冠报名名单，基耶萨和远藤航未被列入。\u003c/li\u003e\n\u003cli\u003e为什么重要:这是一则足球阵容新闻，与人工智能技术、产业或研究没有直接关联，主要体现了热度话题分类可能存在跨领域噪声。\u003c/li\u003e\n\u003cli\u003e讨论概况:X上的讨论焦点主要集中在两名球员落选的原因、球队欧冠阵容取舍及后续出场机会；由于缺少代表性推文，暂无法确认具体舆论倾向。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-8sophie-cunningham-enjoys-fresh-ground-chuck-from-family-friends-ranch\"\u003e\n  话题 8:Sophie Cunningham Enjoys Fresh Ground Chuck from Family Friend\u0026rsquo;s Ranch\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-8sophie-cunningham-enjoys-fresh-ground-chuck-from-family-friends-ranch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Sports\u003c/li\u003e\n\u003cli\u003e概况:热度时间:23 hours ago,相关帖子数:1800\u003c/li\u003e\n\u003cli\u003e是什么事:Sophie Cunningham分享或谈及享用来自家族朋友牧场的新鲜牛肉末。\u003c/li\u003e\n\u003cli\u003e为什么重要:从给定信息看，该话题与AI技术或产业没有直接关联，归入AI分类可能反映了平台热榜的标签偏差或语境识别问题。\u003c/li\u003e\n\u003cli\u003e讨论概况:材料未提供代表推文，无法可靠判断X上的具体讨论焦点；目前可确认的内容主要围绕球员生活与食品来源展开。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-91968-nyc-subway-photos-spark-debate-on-change-and-order\"\u003e\n  话题 9:1968 NYC Subway Photos Spark Debate on Change and Order\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-91968-nyc-subway-photos-spark-debate-on-change-and-order\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Sports\u003c/li\u003e\n\u003cli\u003e概况:热度时间:2 hours ago,相关帖子数:405\u003c/li\u003e\n\u003cli\u003e摘要:1968 NYC Subway Photos Spark Debate on Change and Order:\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-10gabriel-martinelli-joins-al-hilal-in-arsenals-record-70m-sale\"\u003e\n  话题 10:Gabriel Martinelli Joins Al-Hilal in Arsenal\u0026rsquo;s Record €70m Sale\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-10gabriel-martinelli-joins-al-hilal-in-arsenals-record-70m-sale\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Sports\u003c/li\u003e\n\u003cli\u003e概况:热度时间:2 days ago,相关帖子数:192000\u003c/li\u003e\n\u003cli\u003e摘要:Gabriel Martinelli Joins Al-Hilal in Arsenal\u0026rsquo;s Record €70m Sale: 🚨🔵⚪️ OFFICIAL: Gabriel Martinelli joins Al Hilal from Arsenal in a €70m package deal. 🇧🇷 It becomes Arsenal’s record sale, with a sell-on clause also included in the agreement.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-11lisa-reveals-hidden-romances-and-k-pop-struggles-in-new-documentary\"\u003e\n  话题 11:Lisa Reveals Hidden Romances and K-pop Struggles in New Documentary\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-11lisa-reveals-hidden-romances-and-k-pop-struggles-in-new-documentary\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Entertainment\u003c/li\u003e\n\u003cli\u003e概况:热度时间:20 hours ago,相关帖子数:61000\u003c/li\u003e\n\u003cli\u003e摘要:Lisa Reveals Hidden Romances and K-pop Struggles in New Documentary:\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-12elon-musk-documentary-sparks-family-defense-and-fierce-backlash\"\u003e\n  话题 12:Elon Musk Documentary Sparks Family Defense and Fierce Backlash\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-12elon-musk-documentary-sparks-family-defense-and-fierce-backlash\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Entertainment\u003c/li\u003e\n\u003cli\u003e概况:热度时间:1 day ago,相关帖子数:49000\u003c/li\u003e\n\u003cli\u003e是什么事:一部关于埃隆·马斯克的纪录片引发关注后，其家人出面为他辩护，同时也在 X 上招致大量批评和反弹。\u003c/li\u003e\n\u003cli\u003e为什么重要:马斯克同时是 AI 公司 xAI 的核心人物，这类舆论事件会影响外界对其个人形象、商业决策和 AI 版图的判断，也会放大围绕 AI 领导者责任与影响力的讨论。\u003c/li\u003e\n\u003cli\u003e讨论概况:X 上主要在争论纪录片是否公正呈现马斯克、家人辩护是否有说服力，以及马斯克的个人争议是否会连带影响 xAI、特斯拉和相关 AI 议题的公众信任。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-13maye-musk-rejects-claim-of-elons-misery-from-anonymous-source\"\u003e\n  话题 13:Maye Musk Rejects Claim of Elon\u0026rsquo;s Misery from Anonymous Source\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-13maye-musk-rejects-claim-of-elons-misery-from-anonymous-source\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Entertainment\u003c/li\u003e\n\u003cli\u003e概况:热度时间:6 hours ago,相关帖子数:7800\u003c/li\u003e\n\u003cli\u003e是什么事:Maye Musk公开否认了匿名来源关于“埃隆·马斯克很痛苦/不快乐”的说法，回应了围绕其个人状态的传闻。\u003c/li\u003e\n\u003cli\u003e为什么重要:马斯克是 xAI、Tesla 和 X 的核心人物，他的个人形象、情绪状态和公众叙事会直接影响外界对其 AI 业务、管理风格和战略判断的关注。\u003c/li\u003e\n\u003cli\u003e讨论概况:X 上的讨论主要集中在消息来源是否可靠、Maye Musk 的否认是否足以反驳传闻，以及马斯克的个人生活是否会被过度放大并影响对其 AI 事业的评价。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-14elon-musk-warns-austin-heat-hinders-recruiting\"\u003e\n  话题 14:Elon Musk Warns Austin Heat Hinders Recruiting\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-14elon-musk-warns-austin-heat-hinders-recruiting\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · News\u003c/li\u003e\n\u003cli\u003e概况:热度时间:3 hours ago,相关帖子数:1700\u003c/li\u003e\n\u003cli\u003e摘要:Elon Musk Warns Austin Heat Hinders Recruiting:\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"话题-15healthcare-worker-leaves-voicemail-for-joy-as-hope\"\u003e\n  话题 15:Healthcare Worker Leaves Voicemail for Joy as Hope\n  \u003ca class=\"heading-link\" href=\"#%e8%af%9d%e9%a2%98-15healthcare-worker-leaves-voicemail-for-joy-as-hope\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e分类:AI · Entertainment\u003c/li\u003e\n\u003cli\u003e概况:热度时间:3 hours ago,相关帖子数:7000\u003c/li\u003e\n\u003cli\u003e是什么事:一名医护人员给名为 Joy 的对象留言，表达希望，这段内容在 X 上引发关注。\u003c/li\u003e\n\u003cli\u003e为什么重要:该话题体现了 AI 与情感化娱乐内容结合的传播潜力，也引发对合成语音、真实身份和内容真实性的关注。\u003c/li\u003e\n\u003cli\u003e讨论概况:讨论焦点集中在留言是否由 AI 生成或经过声音合成、故事是否真实，以及这种具有情绪感染力的内容是在传递希望还是利用情感博取流量。\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2 id=\"-大佬观点influencer-insights\"\u003e\n  💡 大佬观点(Influencer Insights)\n  \u003ca class=\"heading-link\" href=\"#-%e5%a4%a7%e4%bd%ac%e8%a7%82%e7%82%b9influencer-insights\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003cblockquote\u003e\n\u003cp\u003e今日大佬观点暂缺,推荐阅读 Watch List 深度内容。\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch2 id=\"-附录今日-watch-list-更新源列表\"\u003e\n  📚 附录:今日 Watch List 更新源列表\n  \u003ca class=\"heading-link\" href=\"#-%e9%99%84%e5%bd%95%e4%bb%8a%e6%97%a5-watch-list-%e6%9b%b4%e6%96%b0%e6%ba%90%e5%88%97%e8%a1%a8\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003cblockquote\u003e\n\u003cp\u003e时间窗口:最近 3 天;覆盖 22 个源;共 36 条更新\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch3 id=\"openai-blog-a_full\"\u003e\n  OpenAI Blog (A_full)\n  \u003ca class=\"heading-link\" href=\"#openai-blog-a_full\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://openai.com/index/daybreak-for-frontline-defenders\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eDaybreak for Frontline Defenders: $1B to protect essential services\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 21:15 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- Today OpenAI is introducing Daybreak for Frontline Defenders, a new global initiative to help frontline defenders use frontier AI cyber capabilities to protect essential services in the United States and around the world.\n\u003cul\u003e\n\u003cli\u003e\n\u003cul\u003e\n\u003cli\u003eA $1 billion global commitment to expand subsidized access to Daybreak cyber models and products, training, technical support, and partnerships in the United States and internationally.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cul\u003e\n\u003cli\u003eDaybreak for America, bringing together all of OpenAI’s U.S.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003ework to protect the systems Americans rely on every day—from water and electricity to local government and banking—including a new pilot with the Multi-State Information Sharing and Analysis Center (MS-ISAC).\u003c/li\u003e\n\u003cli\u003e\n\u003cul\u003e\n\u003cli\u003eMore than 35 enterprise products and partner-operated services through the Daybreak Defense Network, bringing Daybreak cyber models into the tools, services, and workflows enterprise defenders already use.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003eOpenAI introduces Daybreak for Frontline Defenders\u003c/li\u003e\n\u003cli\u003eA $1 billion commitment expands access to frontier cyber AI, training, and support for essential services.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://openai.com/index/legora-financial-statement-review-with-astra\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eLegora reviewed 41 documents in minutes with GPT-6 Astra\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 20:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- Legora is an agentic operating system for legal and professional work, used by more than 100,000 professionals across more than 1,800 in-house legal departments and law firms in over 50 markets.\n\u003cul\u003e\n\u003cli\u003eIts legal engineers work directly with customers to understand how they operate and adapt Legora to their end-to-end workflows, from contract and agreement review to legal research.\u003c/li\u003e\n\u003cli\u003eOne of the more tedious workflows is financial-statement tie-out: checking every figure in draft accounts against trial balances, a consolidation schedule, and the previous year’s accounts until each item agrees.\u003c/li\u003e\n\u003cli\u003eAs Legora Legal Engineer Percevale Perks says, the work “can take an entire evening, sometimes days.”.\u003c/li\u003e\n\u003cli\u003e\n\u003ch2 id=\"processing-complex-financial-context-at-scale\"\u003e\n  Processing complex financial context at scale.\n  \u003ca class=\"heading-link\" href=\"#processing-complex-financial-context-at-scale\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003eLegora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://openai.com/index/playco-game-prototyping-with-astra\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePlayco cut manual fixes 50% prototyping games with GPT-6 Astra\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 20:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.\n\u003cul\u003e\n\u003cli\u003eThis piece from OpenAI Blog explains how Playco cut manual fixes 50% prototyping games with GPT-6 Astra shapes the broader AI and infrastructure landscape.\u003c/li\u003e\n\u003cli\u003eIt also surfaces practical implications for founders, operators, and investors following Playco cut manual fixes 50% prototyping games with GPT-6 Astra.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003eUsing GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://openai.com/index/safety-overview-gpt-6-astra\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSafety overview: GPT-6 Astra\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 08:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- GPT-6 Astra is our most capable broadly deployed model and our first to reach the Critical level of cybersecurity capability under our Preparedness Framework.\n\u003cul\u003e\n\u003cli\u003eThis piece from OpenAI Blog explains how Safety overview: GPT-6 Astra shapes the broader AI and infrastructure landscape.\u003c/li\u003e\n\u003cli\u003eIt also surfaces practical implications for founders, operators, and investors following Safety overview: GPT-6 Astra.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003eGPT-6 Astra is our most capable broadly deployed model and our first to reach the Critical level of cybersecurity capability under our Preparedness Framework.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"google-deepmind-blog-a_full\"\u003e\n  Google DeepMind Blog (A_full)\n  \u003ca class=\"heading-link\" href=\"#google-deepmind-blog-a_full\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003e\u003ca href=\"https://deepmind.google/blog/introducing-weathernext-3-our-most-advanced-and-accurate-global-weather-ai-model/\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eIntroducing WeatherNext 3, our most advanced and accurate global weather AI model\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 23:02 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- Introducing WeatherNext 3, our most advanced and accurate global weather AI model.\n\u003cul\u003e\n\u003cli\u003eThis piece from Google DeepMind Blog explains how Introducing WeatherNext 3, our most advanced and accurate global weather AI model shapes the broader AI and infrastructure landscape.\u003c/li\u003e\n\u003cli\u003eIt also surfaces practical implications for founders, operators, and investors following Introducing WeatherNext 3, our most advanced and accurate global weather AI model.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003eIntroducing WeatherNext 3, our most advanced and accurate global weather AI model\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"two-minute-papers-b_introsearch\"\u003e\n  Two Minute Papers (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#two-minute-papers-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003e\u003ca href=\"https://www.youtube.com/watch?v=B3LXEW4Pc-w\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eClaude Fable AI Is Much Stranger Than The Headlines Suggest\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 16:23 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- ❤️ Check out Lambda here and sign up for their GPU Cloud:.\n\u003cul\u003e\n\u003cli\u003e📝 The Claude Fable 5.1 paper is available here:.\u003c/li\u003e\n\u003cli\u003eAdam Bridges, B Shang, Carlos Galarza, Christian Ahlin, Eric Tyson, Juan Benet, Lukas Biewald, Michael Tedder, Owen Skarpness, Ryan Stankye, Shawn Becker, Steef, Taras Bobrovytsky, Tazaur Sagenclaw, Tybie Fitzhugh, Ueli Gallizzi.\u003c/li\u003e\n\u003cli\u003eClaude Fable AI Is Much Stranger Than The Headlines Suggest.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003e❤️ Check out Lambda here and sign up for their GPU Cloud:\u003c/li\u003e\n\u003cli\u003e📝 The Claude Fable 5.1 paper is available here:\u003c/li\u003e\n\u003cli\u003e🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:\u003c/li\u003e\n\u003cli\u003eAdam Bridges, B Shang, Carlos Galarza, Christian Ahlin, Eric Tyson, Juan Benet, Lukas Biewald, Michael Tedder, Owen Skarpness, Ryan Stankye, Shawn Becker, Steef…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"arxiv-csai-b_introsearch\"\u003e\n  ArXiv cs.AI (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#arxiv-csai-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01611\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eEvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01611v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Frontier large language models can often recognize when they are being evaluated, a capability known as evaluation awareness.\u003c/li\u003e\n\u003cli\u003eIf models behave differently in evaluations than in deployment, this undermines the validity of evaluation results, which are a crucial component of current AI safety frameworks.\u003c/li\u003e\n\u003cli\u003eWe introduce EvalDetectBench, an open pipeline and benchmark for measuring evaluation awareness that works with any Inspect-compatible evaluation, allowing practitioners to test against current and future benchmarks.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01611v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Frontier large language models can often recognize when they are being evaluated, a capability known as evaluation awareness\u003c/li\u003e\n\u003cli\u003eIf models behave differently in evaluations than in deployment, this undermines the validity of evaluation results, which are a crucial component of current AI…\u003c/li\u003e\n\u003cli\u003eWe introduce EvalDetectBench, an open pipeline and benchmark for measuring evaluation awareness that works with any Inspect-compatible evaluation, allowing prac…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01685\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eMeta-ethics and AI: exploring the novel meta-ethical questions in the era of AI\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01685v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: With the development of artificial intelligence (AI), the landscape of meta-ethics, which has largely centred on human ethics, faces pressures that may significantly reconfigure it.\u003c/li\u003e\n\u003cli\u003eIn particular, if future AI systems were to exhibit sufficiently integrated capacities for moral reasoning, moral intentionality, and moral reflection, novel meta-ethical questions would arise concerning what I call \u0026ldquo;AI\u0026rsquo;s own ethics\u0026rdquo;, as distinct from ethical principles merely imposed on AI by human designers.\u003c/li\u003e\n\u003cli\u003eThis paper offers a conditional and methodological framework for identifying the questions that would emerge if such AI systems were to arise.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01685v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: With the development of artificial intelligence (AI), the landscape of meta-ethics, which has largely centred on human ethics, faces pressures that ma…\u003c/li\u003e\n\u003cli\u003eIn particular, if future AI systems were to exhibit sufficiently integrated capacities for moral reasoning, moral intentionality, and moral reflection, novel me…\u003c/li\u003e\n\u003cli\u003eThis paper offers a conditional and methodological framework for identifying the questions that would emerge if such AI systems were to arise\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01741\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eWhen Can a Machine Trust a Statute? A Survival Certificate for Machine-Extracted Legal Logic\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01741v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Statutes are increasingly parsed by machines before people read them, and the parsers disagree: on Missouri\u0026rsquo;s statutes, two independently written extractors diverge on numeric-threshold presence at a false-negative rate of 0.43.\u003c/li\u003e\n\u003cli\u003eWe ask what formal logic survives such noise.\u003c/li\u003e\n\u003cli\u003eWe build a passive survival certificate for the Duquenne-Guigues implication basis of machine-extracted statutory contexts: per-attribute inter-extractor disagreement is measured, replayed against the basis in 1,000 Monte Carlo trials, and an implication is certified only when a one-sided Wilson 95% lower bound on survival reaches 0.95; every certified implication carries premise spans and a minimal counterexample.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01741v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Statutes are increasingly parsed by machines before people read them, and the parsers disagree: on Missouri\u0026rsquo;s statutes, two independently written extr…\u003c/li\u003e\n\u003cli\u003eWe ask what formal logic survives such noise\u003c/li\u003e\n\u003cli\u003eWe build a passive survival certificate for the Duquenne-Guigues implication basis of machine-extracted statutory contexts: per-attribute inter-extractor disagr…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01814\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eWhen Does Information Sharing Improve Decentralized Discovery? Aggregation, Independent Rescue, and Equilibrium Selection\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01814v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Information sharing can improve a pooled estimate while eliminating independent rescue actions.\u003c/li\u003e\n\u003cli\u003eThis paper separates those effects in exact finite discovery models.\u003c/li\u003e\n\u003cli\u003eA centralized action-budget profile shows that equal one-person accuracy can coexist with different portfolio values.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01814v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Information sharing can improve a pooled estimate while eliminating independent rescue actions\u003c/li\u003e\n\u003cli\u003eThis paper separates those effects in exact finite discovery models\u003c/li\u003e\n\u003cli\u003eA centralized action-budget profile shows that equal one-person accuracy can coexist with different portfolio values\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01815\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eInduction and Inquiry via Probabilistic Reasoning over Language and Code\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01815v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: How humans grow and maintain abstract knowledge from the sparse, streaming noisy data of experience is a longstanding challenge in cognitive science.\u003c/li\u003e\n\u003cli\u003eAny computational account must satisfy at least three desiderata: It must be (1) data-efficient and compute-efficient, (2) capture gradations of uncertainty to support intelligent inquiry and information gathering, and (3) be flexible enough to mentally represent the endless range of concepts people can learn and think about.\u003c/li\u003e\n\u003cli\u003eHere we introduce a computational model that captures these three properties, by encoding symbolic knowledge as mental programs that combine natural language with source code, and sequentially inferring mental programs using LLM-guided Bayesian learning algorithms.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01815v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: How humans grow and maintain abstract knowledge from the sparse, streaming noisy data of experience is a longstanding challenge in cognitive science\u003c/li\u003e\n\u003cli\u003eAny computational account must satisfy at least three desiderata: It must be (1) data-efficient and compute-efficient, (2) capture gradations of uncertainty to…\u003c/li\u003e\n\u003cli\u003eHere we introduce a computational model that captures these three properties, by encoding symbolic knowledge as mental programs that combine natural language wi…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01834\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eArchitecting Conversational Data Systems for Stateless LLM APIs: The Hydration Proxy Pattern\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01834v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: As enterprise platforms transition to conversational reasoning interfaces, the stateless nature of LLM APIs creates an architectural gap.\u003c/li\u003e\n\u003cli\u003eWhile statelessness enables horizontal scalability for AI providers, it forces client applications to manage the entire burden of conversational state and semantic memory.\u003c/li\u003e\n\u003cli\u003eThe work identifies the Hydration Proxy Pattern, an architecture that decouples session persistence from the reasoning engine.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01834v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: As enterprise platforms transition to conversational reasoning interfaces, the stateless nature of LLM APIs creates an architectural gap\u003c/li\u003e\n\u003cli\u003eWhile statelessness enables horizontal scalability for AI providers, it forces client applications to manage the entire burden of conversational state and seman…\u003c/li\u003e\n\u003cli\u003eThe work identifies the Hydration Proxy Pattern, an architecture that decouples session persistence from the reasoning engine\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01849\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSSAKG 2.0: An Open-Source Package for Structural Associative Sequence Memory and Context-Based Retrieval\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01849v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: This article presents SSAKG 2.0, an open-source software package for constructing and operating Structural Sequential Associative Knowledge Graphs (SSAKGs).\u003c/li\u003e\n\u003cli\u003eAn SSAKG represents objects as graph vertices and ordered sequences as structural patterns of graph connections.\u003c/li\u003e\n\u003cli\u003eThe resulting sparse graph is used as an associative memory in which complete sequences can be reconstructed from a partial, unordered context.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01849v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: This article presents SSAKG 2.0, an open-source software package for constructing and operating Structural Sequential Associative Knowledge Graphs (SS…\u003c/li\u003e\n\u003cli\u003eAn SSAKG represents objects as graph vertices and ordered sequences as structural patterns of graph connections\u003c/li\u003e\n\u003cli\u003eThe resulting sparse graph is used as an associative memory in which complete sequences can be reconstructed from a partial, unordered context\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01852\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eThe Memory Trust Gap: Capability-Dependent Failures in Persistent-Memory Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01852v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Persistent memory supports personalized agents, but a stale stored fact can override current authoritative evidence without warning.\u003c/li\u003e\n\u003cli\u003eWe study when this harm begins as model capability changes.\u003c/li\u003e\n\u003cli\u003eWe evaluate a frozen, closed-set, action-scored benchmark with 2 suites that represent 2 different meanings of \u0026ldquo;no memory\u0026rdquo; (a Benefit suite, unsolvable without the stored fact, and a Safety suite, in which an authoritative tool always holds the correct value), on a same-family model-size series (Qwen3 0.6/1.7/4/8B).\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01852v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Persistent memory supports personalized agents, but a stale stored fact can override current authoritative evidence without warning\u003c/li\u003e\n\u003cli\u003eWe study when this harm begins as model capability changes\u003c/li\u003e\n\u003cli\u003eWe evaluate a frozen, closed-set, action-scored benchmark with 2 suites that represent 2 different meanings of \u0026ldquo;no memory\u0026rdquo; (a Benefit suite, unsolvable without…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01861\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eBelief-Calibrated Optimization: An Explicit World Model for Agentic Optimization\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01861v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: The performance of an LLM agent depends on the scaffold around a frozen model.\u003c/li\u003e\n\u003cli\u003eA common way to improve that scaffold is to use a coding agent as an optimizer: it reads current scores and traces and iteratively edits the source, producing a new candidate each round.\u003c/li\u003e\n\u003cli\u003eEach edit is chosen according to a belief about how the environment will respond: what went wrong, and which change should help.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01861v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: The performance of an LLM agent depends on the scaffold around a frozen model\u003c/li\u003e\n\u003cli\u003eA common way to improve that scaffold is to use a coding agent as an optimizer: it reads current scores and traces and iteratively edits the source, producing a…\u003c/li\u003e\n\u003cli\u003eEach edit is chosen according to a belief about how the environment will respond: what went wrong, and which change should help\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01873\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eEpistemic Sybil Resistance: Multiplying AI Agents Without Multiplying Evidence\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01873v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Multi-agent AI systems improve inference by spawning agents and synthesizing reports.\u003c/li\u003e\n\u003cli\u003eBut another agent is not another observation: apparently independent reports may descend from the same evidence, and genuinely independent evidence can produce nearly identical reports.\u003c/li\u003e\n\u003cli\u003eWe formalize this as an epistemic Sybil problem.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01873v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Multi-agent AI systems improve inference by spawning agents and synthesizing reports\u003c/li\u003e\n\u003cli\u003eBut another agent is not another observation: apparently independent reports may descend from the same evidence, and genuinely independent evidence can produce…\u003c/li\u003e\n\u003cli\u003eWe formalize this as an epistemic Sybil problem\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"arxiv-cscl-b_introsearch\"\u003e\n  ArXiv cs.CL (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#arxiv-cscl-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01658\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePRO-Step: Step-level Process Reward Optimization for Retrieval-Augmented Generation\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01658v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Retrieval-Augmented Generation enhances Large Language Models by grounding responses in external knowledge, but multi-hop reasoning remains vulnerable to error propagation, where early retrieval failures confound subsequent steps.\u003c/li\u003e\n\u003cli\u003eStandard outcome-based optimization only rewards the final answer, leaving intermediate retrieval and reasoning errors undetected.\u003c/li\u003e\n\u003cli\u003eWhile existing process-based methods introduce step-level signals, they still score each step against the final answer, rewarding spurious successes where flawed retrieval coincidentally produces the correct answer.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01658v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Retrieval-Augmented Generation enhances Large Language Models by grounding responses in external knowledge, but multi-hop reasoning remains vulnerable…\u003c/li\u003e\n\u003cli\u003eStandard outcome-based optimization only rewards the final answer, leaving intermediate retrieval and reasoning errors undetected\u003c/li\u003e\n\u003cli\u003eWhile existing process-based methods introduce step-level signals, they still score each step against the final answer, rewarding spurious successes where flawe…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01687\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eLearning Evidence Sufficiency Boundaries for Selective Answering in Grounded Multi-Hop QA\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01687v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Grounded question answering systems should answer only when the supplied evidence supports the answer.\u003c/li\u003e\n\u003cli\u003eIn multi-hop QA, this requirement is difficult because partial evidence can make an unsupported answer appear plausible.\u003c/li\u003e\n\u003cli\u003eWe study selective answering through evidence sufficiency boundaries: for the same question, a model should abstain under unsupported or partially supported context, answer when the context first becomes sufficient, and keep the answer stable when redundant evidence is added.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01687v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Grounded question answering systems should answer only when the supplied evidence supports the answer\u003c/li\u003e\n\u003cli\u003eIn multi-hop QA, this requirement is difficult because partial evidence can make an unsupported answer appear plausible\u003c/li\u003e\n\u003cli\u003eWe study selective answering through evidence sufficiency boundaries: for the same question, a model should abstain under unsupported or partially supported con…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01737\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSpeakPay: Domain-Adaptive LoRA Fine-Tuning of Whisper for Low-Resource Nepali Financial Speech Recognition\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01737v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Mobile payment applications in Nepal are graphically mediated and largely inaccessible to visually impaired users.\u003c/li\u003e\n\u003cli\u003eThis paper presents SpeakPay, a voice-first digital wallet, and documents the central technical contribution: a controlled study of domain adaptation for low-resource financial speech recognition.\u003c/li\u003e\n\u003cli\u003eWe introduce NepFinSpeech-403, a 403-utterance dataset of Nepali financial voice commands (send, load, and balance operations spanning 237 unique numerals), and fine-tune Whisper large-v2 with LoRA.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01737v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Mobile payment applications in Nepal are graphically mediated and largely inaccessible to visually impaired users\u003c/li\u003e\n\u003cli\u003eThis paper presents SpeakPay, a voice-first digital wallet, and documents the central technical contribution: a controlled study of domain adaptation for low-re…\u003c/li\u003e\n\u003cli\u003eWe introduce NepFinSpeech-403, a 403-utterance dataset of Nepali financial voice commands (send, load, and balance operations spanning 237 unique numerals), and…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01772\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eMemeCULT-1K: Benchmarking South Asian Cultural Context and Humor Understanding of Multimodal Models\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01772v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Meme understanding goes beyond recognizing visual content or literal text; it requires implicit cultural knowledge and pragmatic inference that most vision-language models still lack.\u003c/li\u003e\n\u003cli\u003eWe introduce MemeCULT-1K, a multilingual benchmark of 1,000 South Asian memes in Bengali, English, and Hindi, where each meme is paired with a cultural context note and three human-written explanations, along with a supplementary set of 54 Bengali regional dialect memes.\u003c/li\u003e\n\u003cli\u003eWe evaluate thirteen popular Vision Language Models (VLMs) under two settings: meme-only and context-aware.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01772v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Meme understanding goes beyond recognizing visual content or literal text; it requires implicit cultural knowledge and pragmatic inference that most v…\u003c/li\u003e\n\u003cli\u003eWe introduce MemeCULT-1K, a multilingual benchmark of 1,000 South Asian memes in Bengali, English, and Hindi, where each meme is paired with a cultural context…\u003c/li\u003e\n\u003cli\u003eWe evaluate thirteen popular Vision Language Models (VLMs) under two settings: meme-only and context-aware\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01788\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eVakyArth: Evaluating Pragmatic Competence in LLMs across Indic Languages\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01788v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Real-world communication often requires pragmatic reasoning: interpreting meanings implied through context and cultural convention rather than stated literally.\u003c/li\u003e\n\u003cli\u003eExisting pragmatic evaluation remains largely limited to English and high-resource languages, leaving Indic languages unexplored despite their linguistic and cultural diversity.\u003c/li\u003e\n\u003cli\u003eWe introduce VakyArth, the first pragmatic benchmark for Indic languages, designed as a diagnostic evaluation covering Hindi, Punjabi, Tamil, and Malayalam.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01788v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Real-world communication often requires pragmatic reasoning: interpreting meanings implied through context and cultural convention rather than stated…\u003c/li\u003e\n\u003cli\u003eExisting pragmatic evaluation remains largely limited to English and high-resource languages, leaving Indic languages unexplored despite their linguistic and cu…\u003c/li\u003e\n\u003cli\u003eWe introduce VakyArth, the first pragmatic benchmark for Indic languages, designed as a diagnostic evaluation covering Hindi, Punjabi, Tamil, and Malayalam\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01794\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eDisentangling Statistical Preemption from Entrenchment in Language Models\u0026rsquo; Avoidance of Overgeneralization\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01794v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: How do learners avoid overgeneralizations such as Tom laughed me without explicit negative evidence?\u003c/li\u003e\n\u003cli\u003eConstructionists have posited two proposals that describe indirect negative evidence against overgeneralizations: preemption (which privileges exposure to near-synonymous construction\u0026mdash;e.g., she made him laugh) vs.\u003c/li\u003e\n\u003cli\u003eentrenchment (all exposures to a verb\u0026rsquo;s grammatical usages, including cases like He laughed).\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01794v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: How do learners avoid overgeneralizations such as Tom laughed me without explicit negative evidence\u003c/li\u003e\n\u003cli\u003eConstructionists have posited two proposals that describe indirect negative evidence against overgeneralizations: preemption (which privileges exposure to near-…\u003c/li\u003e\n\u003cli\u003eentrenchment (all exposures to a verb\u0026rsquo;s grammatical usages, including cases like He laughed)\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01798\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eHow Do Prompt Variations Affect Energy Consumption in On-Device LLMs?\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01798v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Large language models (LLMs) are increasingly deployed on mobile devices, making energy efficiency a key deployment constraint, yet the energy impact of prompt design remains underexplored.\u003c/li\u003e\n\u003cli\u003eThis paper aims to understand how two prompt properties, cognitive load and phrasing pattern, shape the energy behavior of on-device LLM inference.\u003c/li\u003e\n\u003cli\u003eWe conduct a broad empirical study covering prompt properties, datasets, models, and devices, with phase-level profiling that separates prefill and decode energy.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01798v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Large language models (LLMs) are increasingly deployed on mobile devices, making energy efficiency a key deployment constraint, yet the energy impact…\u003c/li\u003e\n\u003cli\u003eThis paper aims to understand how two prompt properties, cognitive load and phrasing pattern, shape the energy behavior of on-device LLM inference\u003c/li\u003e\n\u003cli\u003eWe conduct a broad empirical study covering prompt properties, datasets, models, and devices, with phase-level profiling that separates prefill and decode energ…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01810\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eTalkFa: A Unified Benchmark for Farsi Dialogue Generation and Understanding\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01810v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Farsi, spoken by more than 120 million people, lacks a comprehensive benchmark for dialogue generation and understanding.\u003c/li\u003e\n\u003cli\u003eWe introduce TALKFA, a unified benchmark comprising three complementary datasets: (1) WIKI-FADIAL, 4.2K Wikipedia-grounded dialogues for knowledge-grounded generation; (2) DAILYDIALOG-FA, 6.6K dialogues annotated for dialogue acts and emotions; and (3) PLAYDIAL-FA, 2.1K theatrical dialogues with sentiment labels.\u003c/li\u003e\n\u003cli\u003eWhile LLMs assist data construction, every dialogue undergoes multi-stage review and revision by native Farsi speakers, and only the final human-approved dialogues are released.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01810v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Farsi, spoken by more than 120 million people, lacks a comprehensive benchmark for dialogue generation and understanding\u003c/li\u003e\n\u003cli\u003eWe introduce TALKFA, a unified benchmark comprising three complementary datasets: (1) WIKI-FADIAL, 4.2K Wikipedia-grounded dialogues for knowledge-grounded gene…\u003c/li\u003e\n\u003cli\u003eWhile LLMs assist data construction, every dialogue undergoes multi-stage review and revision by native Farsi speakers, and only the final human-approved dialog…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01828\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eAVERT: Audio-Verified Adjudication for Spoken Dialogue State Tracking\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01828v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Spoken dialogue state tracking recovers slot-value pairs from speech, where ASR errors concentrate in entity values and persist across turns, making it both a generation and an editing problem.\u003c/li\u003e\n\u003cli\u003eA strong per-turn text editor corrects much of this but, operating on the transcript alone, leaves three recoverable errors: a value predicted inconsistently across turns, an omitted slot, and a value the audio does not support.\u003c/li\u003e\n\u003cli\u003eWe present AVERT, which scores each candidate value by combining cross-turn agreement with a trained audio-conditioned verifier and resolves the three error types with three operators, vote, add, and swap, each restricted to the slots where its error is common.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01828v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Spoken dialogue state tracking recovers slot-value pairs from speech, where ASR errors concentrate in entity values and persist across turns, making i…\u003c/li\u003e\n\u003cli\u003eA strong per-turn text editor corrects much of this but, operating on the transcript alone, leaves three recoverable errors: a value predicted inconsistently ac…\u003c/li\u003e\n\u003cli\u003eWe present AVERT, which scores each candidate value by combining cross-turn agreement with a trained audio-conditioned verifier and resolves the three error typ…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01832\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eInterpretable Symptom Vectors for Depression in a Large Language Model\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01832v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Patients with depression present with diverse symptom profiles, yet clinical practice routinely reduces this variation to a single severity score.\u003c/li\u003e\n\u003cli\u003eLarge language models (LLMs) can potentially capture various symptoms and their severity from patient speech.\u003c/li\u003e\n\u003cli\u003eHowever, how depressive symptoms are represented inside LLMs remains poorly understood, limiting clinical trust.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01832v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Patients with depression present with diverse symptom profiles, yet clinical practice routinely reduces this variation to a single severity score\u003c/li\u003e\n\u003cli\u003eLarge language models (LLMs) can potentially capture various symptoms and their severity from patient speech\u003c/li\u003e\n\u003cli\u003eHowever, how depressive symptoms are represented inside LLMs remains poorly understood, limiting clinical trust\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"arxiv-cslg-b_introsearch\"\u003e\n  ArXiv cs.LG (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#arxiv-cslg-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"链接到标题\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003e链接到标题\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01608\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eWMLLM: Self-Evolving Optimization Agents via Predict-Then-Act World Modeling\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01608v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Black-box optimization problems remain challenging because of large, weakly structured, and high-dimensional search spaces.\u003c/li\u003e\n\u003cli\u003eExisting methods often suffer from poor sample efficiency because they rely on direct candidate generation or trial-and-error refinement.\u003c/li\u003e\n\u003cli\u003eA natural way to improve search efficiency is to use world modeling, which can help identify promising optimization directions before costly evaluation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01608v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Black-box optimization problems remain challenging because of large, weakly structured, and high-dimensional search spaces\u003c/li\u003e\n\u003cli\u003eExisting methods often suffer from poor sample efficiency because they rely on direct candidate generation or trial-and-error refinement\u003c/li\u003e\n\u003cli\u003eA natural way to improve search efficiency is to use world modeling, which can help identify promising optimization directions before costly evaluation\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01609\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eDiDrive: A Risk-Aware Hierarchical Diffusion Framework for Safe Offline Reinforcement Learning in Autonomous Driving\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01609v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: While diffusion models effectively capture multimodal behavioral priors for autonomous driving, offline reinforcement learning (RL) policies remain susceptible to distribution shift, heavy-tailed risk signals, out-of-distribution (OOD) action generation, and high-dimensional state redundancy.\u003c/li\u003e\n\u003cli\u003eTo address these challenges, we propose DiDrive, a distribution-guided offline diffusion framework featuring two synergistic components: the Risk-Aware Hierarchical Diffusion (RHDif) architecture and the 3DICE policy optimization paradigm.\u003c/li\u003e\n\u003cli\u003eIn the state space, RHDif utilizes a low-level risk-gated encoder and a high-level contextual modulator to filter environmental redundancy and focus on safety-critical threats.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01609v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: While diffusion models effectively capture multimodal behavioral priors for autonomous driving, offline reinforcement learning (RL) policies remain su…\u003c/li\u003e\n\u003cli\u003eTo address these challenges, we propose DiDrive, a distribution-guided offline diffusion framework featuring two synergistic components: the Risk-Aware Hierarch…\u003c/li\u003e\n\u003cli\u003eIn the state space, RHDif utilizes a low-level risk-gated encoder and a high-level contextual modulator to filter environmental redundancy and focus on safety-c…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01615\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePrompt-Space Meta-Learning Does Not Transfer Across Users: A Frozen-LLM Negative Result\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01615v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Personalizing a frozen large language model (LLM) to individual users is often framed as a meta-learning problem in prompt space: each user is a task, and one seeks a shared natural-language adaptation policy that, given a handful of the user\u0026rsquo;s labeled interactions, configures the frozen model for that user.\u003c/li\u003e\n\u003cli\u003eThe framing is attractive because it is backbone-agnostic and reuses the machinery of prompt optimization, yet the field rarely tests whether the optimized meta-objective encodes transferable cross-user adaptation rather than generic instruction quality.\u003c/li\u003e\n\u003cli\u003eWe study this question with Muse (Meta-learned User-adaptation via Shared Evolution), which evolves a single shared adaptation prompt over a meta-train user population by reflective prompt evolution, freezes it, and applies it zero-shot to held-out users; matched controls isolate learning from confounds of phrasing and selection.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01615v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Personalizing a frozen large language model (LLM) to individual users is often framed as a meta-learning problem in prompt space: each user is a task,…\u003c/li\u003e\n\u003cli\u003eThe framing is attractive because it is backbone-agnostic and reuses the machinery of prompt optimization, yet the field rarely tests whether the optimized meta…\u003c/li\u003e\n\u003cli\u003eWe study this question with Muse (Meta-learned User-adaptation via Shared Evolution), which evolves a single shared adaptation prompt over a meta-train user pop…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01647\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eEfficient Context-Limited Telescope Bibliography Classification for the WASP-2025 Shared Task Using SciBERT\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01647v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: The creation of telescope bibliographies is a crucial part of assessing the scientific impact of observatories and ensuring reproducibility in astronomy.\u003c/li\u003e\n\u003cli\u003eThis task involves identifying, categorizing, and linking scientific publications that reference or use specific telescopes.\u003c/li\u003e\n\u003cli\u003eHowever, this process remains largely manual and resource intensive.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01647v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: The creation of telescope bibliographies is a crucial part of assessing the scientific impact of observatories and ensuring reproducibility in astrono…\u003c/li\u003e\n\u003cli\u003eThis task involves identifying, categorizing, and linking scientific publications that reference or use specific telescopes\u003c/li\u003e\n\u003cli\u003eHowever, this process remains largely manual and resource intensive\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01673\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eCliffRank: A Dual-Branch Framework for Activity-Cliff Ranking Prediction\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01673v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Activity-cliff ranking remains difficult because local structural changes can cause large activity differences, while high-quality data that resolve the underlying mechanisms remain limited.\u003c/li\u003e\n\u003cli\u003eTo use available activity labels more effectively, we combine absolute-activity regression with ranking-consistency learning.\u003c/li\u003e\n\u003cli\u003eCliffRank trains two parallel predictors with mean squared error, a thresholded listwise loss, and Pairwise Preference Consistency (PPC), which aligns relative ordering in the preference-probability space.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01673v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Activity-cliff ranking remains difficult because local structural changes can cause large activity differences, while high-quality data that resolve t…\u003c/li\u003e\n\u003cli\u003eTo use available activity labels more effectively, we combine absolute-activity regression with ranking-consistency learning\u003c/li\u003e\n\u003cli\u003eCliffRank trains two parallel predictors with mean squared error, a thresholded listwise loss, and Pairwise Preference Consistency (PPC), which aligns relative…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01676\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSim2Signal: Sim-to-Real Benchmarks for Traffic Signal Control\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01676v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Reinforcement learning achieves strong traffic signal control performance in simulation, yet policies trained in simulators often fail once deployed in the real world, a failure known as the Sim-to-Real gap.\u003c/li\u003e\n\u003cli\u003eWhen RL is applied to traffic signal control, this gap arises from several sources: sensing, action execution, traffic dynamics, and the control objective.\u003c/li\u003e\n\u003cli\u003eTheir relative impact and the reliability of existing Sim-to-Real mitigation methods remain insufficiently understood, and the field lacks a standard benchmark for systematically measuring the gap and evaluating mitigation methods.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01676v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Reinforcement learning achieves strong traffic signal control performance in simulation, yet policies trained in simulators often fail once deployed i…\u003c/li\u003e\n\u003cli\u003eWhen RL is applied to traffic signal control, this gap arises from several sources: sensing, action execution, traffic dynamics, and the control objective\u003c/li\u003e\n\u003cli\u003eTheir relative impact and the reliability of existing Sim-to-Real mitigation methods remain insufficiently understood, and the field lacks a standard benchmark…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01679\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eA Survey on Self-Improving Test-Time Intelligence: Feedback-Driven Adapting, Learning, and Scaling at Inference\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01679v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: The ability of AI systems to improve their behavior during deployment is becoming increasingly important.\u003c/li\u003e\n\u003cli\u003eAs inference moves beyond the static execution of a fixed trained model, a growing body of work studies how models can refine their behavior on the fly by exploiting test-time information and additional computation.\u003c/li\u003e\n\u003cli\u003eThese developments have largely evolved along two directions: methods that modify the model\u0026rsquo;s state using test-time signals, and methods that improve predictions through extra inference-time resources such as more sampling and tool use.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01679v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: The ability of AI systems to improve their behavior during deployment is becoming increasingly important\u003c/li\u003e\n\u003cli\u003eAs inference moves beyond the static execution of a fixed trained model, a growing body of work studies how models can refine their behavior on the fly by explo…\u003c/li\u003e\n\u003cli\u003eThese developments have largely evolved along two directions: methods that modify the model\u0026rsquo;s state using test-time signals, and methods that improve prediction…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01680\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eReinforcement Learning and Rule-Based Peer-to-Peer Pricing in Residential PV-BES Communities\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01680v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: This paper compares rule-based and learning-based pricing mechanisms for peer-to-peer (P2P) electricity trading in residential photovoltaic communities.\u003c/li\u003e\n\u003cli\u003eThe rule-based benchmarks comprise bill-sharing as an ex post allocation mechanism, the mid-market rate, and supply-demand-ratio pricing.\u003c/li\u003e\n\u003cli\u003eThe reinforcement-learning (RL) formulation is implemented through a Deep Q-Network and evaluated under multiplier-based and learnable SDR-shaped pricing, with a fixed-parameter SDR variant as a non-learning control.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01680v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: This paper compares rule-based and learning-based pricing mechanisms for peer-to-peer (P2P) electricity trading in residential photovoltaic communitie…\u003c/li\u003e\n\u003cli\u003eThe rule-based benchmarks comprise bill-sharing as an ex post allocation mechanism, the mid-market rate, and supply-demand-ratio pricing\u003c/li\u003e\n\u003cli\u003eThe reinforcement-learning (RL) formulation is implemented through a Deep Q-Network and evaluated under multiplier-based and learnable SDR-shaped pricing, with…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01689\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eMedian-of-Means as an Extremal Convex Estimator and a Nonconvex Route to the Trimmed Oracle\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01689v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: We revisit median-of-means estimation from a deterministic optimization viewpoint and develop a family of block-Lp estimators for robust learning with heavy-tailed and adversarially corrupted data.\u003c/li\u003e\n\u003cli\u003eIn a block contamination model with at least a fraction 1 minus epsilon of good blocks, we first show that every convex block M-estimator has worst-case robustness constant at least 1 divided by 1 minus 2 epsilon.\u003c/li\u003e\n\u003cli\u003eThis matches the classical median-of-means bound and proves that the trimmed-block oracle constant 1 divided by 1 minus epsilon cannot be attained within the convex class.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01689v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: We revisit median-of-means estimation from a deterministic optimization viewpoint and develop a family of block-Lp estimators for robust learning with…\u003c/li\u003e\n\u003cli\u003eIn a block contamination model with at least a fraction 1 minus epsilon of good blocks, we first show that every convex block M-estimator has worst-case robustn…\u003c/li\u003e\n\u003cli\u003eThis matches the classical median-of-means bound and proves that the trimmed-block oracle constant 1 divided by 1 minus epsilon cannot be attained within the co…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2609.01699\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eTri-Band Channel Measurement-Enabled Multi-Layer Digital Twin for Terahertz Wireless Data Centers\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e发布时间:2026-09-03 12:00 北京时间\u003c/li\u003e\n\u003cli\u003e摘要:【待翻译】- arXiv:2609.01699v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: The rapid growth of AI computing has driven increasing demands for flexible and high-capacity data-center interconnections.\u003c/li\u003e\n\u003cli\u003eOwing to its ultra-wide bandwidth and high spatial reuse capability, terahertz (THz) communication has emerged as a promising solution for future wireless data centers, while digital twins (DTs) enable efficient wireless planning and real-time optimization.\u003c/li\u003e\n\u003cli\u003eIn this work, a measurement-driven multi-layer DT framework is proposed for THz wireless data centers, where the physical, channel, evaluation, and manipulation layers are progressively constructed from bottom to top.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2609.01699v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: The rapid growth of AI computing has driven increasing demands for flexible and high-capacity data-center interconnections\u003c/li\u003e\n\u003cli\u003eOwing to its ultra-wide bandwidth and high spatial reuse capability, terahertz (THz) communication has emerged as a promising solution for future wireless data…\u003c/li\u003e\n\u003cli\u003eIn this work, a measurement-driven multi-layer DT framework is proposed for THz wireless data centers, where the physical, channel, evaluation, and manipulation…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n",
  "wordCount": 5968,
  "readingTime": 29,
  "tableOfContents": "\u003cnav id=\"TableOfContents\"\u003e\n  \u003cul\u003e\n    \u003cli\u003e\u003ca href=\"#-本期-watch-list-深度导读\"\u003e📖 本期 Watch List 深度导读\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-x-平台-ai-热点快讯\"\u003e🌐 X 平台 AI 热点快讯\u003c/a\u003e\n      \u003cul\u003e\n        \u003cli\u003e\u003ca href=\"#话题-1openai-launches-gpt-6-astra-as-most-capable-model-yet\"\u003e话题 1:OpenAI Launches GPT-6 Astra as Most Capable Model Yet\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-2lululemon-shares-plunge-15-after-weak-earnings-and-cut-outlook\"\u003e话题 2:Lululemon Shares Plunge 15% After Weak Earnings and Cut Outlook\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-3jd-vance-blames-iran-attacks-for-high-gas-prices-in-white-house-briefing\"\u003e话题 3:JD Vance Blames Iran Attacks for High Gas Prices in White House Briefing\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-4john-ternus-takes-over-as-apples-new-ceo-from-tim-cook\"\u003e话题 4:John Ternus Takes Over as Apple\u0026rsquo;s New CEO from Tim Cook\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-5jd-vance-details-trucking-fraud-crackdown-shutting-down-2000-cdl-mills\"\u003e话题 5:JD Vance Details Trucking Fraud Crackdown Shutting Down 2,000 CDL Mills\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-6nba-hits-clippers-with-historic-penalties-in-kawhi-leonard-cap-probe\"\u003e话题 6:NBA Hits Clippers with Historic Penalties in Kawhi Leonard Cap Probe\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-7liverpool-names-25-man-champions-league-squad-omits-chiesa-and-endo\"\u003e话题 7:Liverpool Names 25-Man Champions League Squad, Omits Chiesa and Endo\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-8sophie-cunningham-enjoys-fresh-ground-chuck-from-family-friends-ranch\"\u003e话题 8:Sophie Cunningham Enjoys Fresh Ground Chuck from Family Friend\u0026rsquo;s Ranch\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-91968-nyc-subway-photos-spark-debate-on-change-and-order\"\u003e话题 9:1968 NYC Subway Photos Spark Debate on Change and Order\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-10gabriel-martinelli-joins-al-hilal-in-arsenals-record-70m-sale\"\u003e话题 10:Gabriel Martinelli Joins Al-Hilal in Arsenal\u0026rsquo;s Record €70m Sale\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-11lisa-reveals-hidden-romances-and-k-pop-struggles-in-new-documentary\"\u003e话题 11:Lisa Reveals Hidden Romances and K-pop Struggles in New Documentary\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-12elon-musk-documentary-sparks-family-defense-and-fierce-backlash\"\u003e话题 12:Elon Musk Documentary Sparks Family Defense and Fierce Backlash\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-13maye-musk-rejects-claim-of-elons-misery-from-anonymous-source\"\u003e话题 13:Maye Musk Rejects Claim of Elon\u0026rsquo;s Misery from Anonymous Source\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-14elon-musk-warns-austin-heat-hinders-recruiting\"\u003e话题 14:Elon Musk Warns Austin Heat Hinders Recruiting\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#话题-15healthcare-worker-leaves-voicemail-for-joy-as-hope\"\u003e话题 15:Healthcare Worker Leaves Voicemail for Joy as Hope\u003c/a\u003e\u003c/li\u003e\n      \u003c/ul\u003e\n    \u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-大佬观点influencer-insights\"\u003e💡 大佬观点(Influencer Insights)\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-附录今日-watch-list-更新源列表\"\u003e📚 附录:今日 Watch List 更新源列表\u003c/a\u003e\n      \u003cul\u003e\n        \u003cli\u003e\u003ca href=\"#openai-blog-a_full\"\u003eOpenAI Blog (A_full)\u003c/a\u003e\u003c/li\u003e\n      \u003c/ul\u003e\n    \u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#processing-complex-financial-context-at-scale\"\u003eProcessing complex financial context at scale.\u003c/a\u003e\n      \u003cul\u003e\n        \u003cli\u003e\u003ca href=\"#google-deepmind-blog-a_full\"\u003eGoogle DeepMind Blog (A_full)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#two-minute-papers-b_introsearch\"\u003eTwo Minute Papers (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#arxiv-csai-b_introsearch\"\u003eArXiv cs.AI (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#arxiv-cscl-b_introsearch\"\u003eArXiv cs.CL (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#arxiv-cslg-b_introsearch\"\u003eArXiv cs.LG (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n      \u003c/ul\u003e\n    \u003c/li\u003e\n  \u003c/ul\u003e\n\u003c/nav\u003e",
  "isDraft": false
}