🤖 AI 速览

今日 AI 的落地边界继续外扩：一方面，连接企业数据、服务金融研究的专业化产品正在形成；另一方面，Codex 与 ChatGPT 已被用于搜索抗菌分子，推动 AI 从信息处理走向科学发现。同时，研究轨迹、结果可靠性与数据权限，正成为规模化应用必须面对的基础问题。
📋 文章元数据
发布时间
2026-09-11
类型
ai-daily
字数
5668
阅读时长
27 min

2026-09-11 AI日更 | 从金融研究到抗菌药物发现，AI加速进入高价值决策链 链接到标题

今日 AI 的落地边界继续外扩：一方面，连接企业数据、服务金融研究的专业化产品正在形成；另一方面，Codex 与 ChatGPT 已被用于搜索抗菌分子，推动 AI 从信息处理走向科学发现。同时，研究轨迹、结果可靠性与数据权限，正成为规模化应用必须面对的基础问题。

📖 本期 Watch List 深度导读 链接到标题

今天最值得关注的，是 AI 正从“通用助手”加速进入高价值工作流：OpenAI 推出面向金融服务的 ChatGPT，并扩大政府网络防御支持；同时，研究者已借助 Codex 和 ChatGPT 搜索抗菌分子，说明 AI 的价值正从问答转向研究与决策。第二条主线是可靠性与可审计性：多篇论文检验模型在跨语言事实拒答、错误数据处理、深度研究和医疗分诊中的表现，值得关注 AI 如何在真实场景中被验证。第三条主线聚焦基础设施，X-CoSD 与 Osprey 分别从协同推测解码和草稿模型训练入手，探索更低成本的推理加速。

🌐 X 平台 AI 热点快讯 链接到标题

话题 1:DeepSeek Launches V4.1 Flash: Faster AI at Lower Cost 链接到标题

  • 分类:AI · News
  • 概况:热度时间:2 days ago,相关帖子数:47000
  • 是什么事:DeepSeek发布V4.1 Flash，主打更快的响应速度和更低的使用成本。
  • 为什么重要:更高的推理效率和更低的成本可能扩大生成式AI的应用范围，并加剧模型服务市场的竞争。
  • 讨论概况:X上的讨论主要聚焦于其实际速度与成本优势、模型性能是否保持稳定，以及它与其他高性价比AI模型的竞争表现。

话题 2:OpenAI Launches Data Agent for ChatGPT Work 链接到标题

  • 分类:AI · News
  • 概况:热度时间:8 hours ago,相关帖子数:1900
  • 是什么事:OpenAI 正在向 ChatGPT Work 用户推出 Data Agent，可连接多种企业数据源，回答分析问题、生成仪表盘并跨工具执行操作。
  • 为什么重要:这标志着企业级 AI 助手从生成内容进一步转向直接访问和操作分散的业务数据，有望降低数据整合成本并推动代理式分析落地。
  • 讨论概况:X 上的讨论主要集中在其支持的数据源范围、无需集中迁移数据的便利性，以及企业数据权限、安全和操作风险；部分用户质疑 OpenAI 未同步公布基准测试，难以独立评估其分析准确性和实际性能。

话题 3:OpenAI Launches ChatGPT for Financial Services Targeting Bankers 链接到标题

  • 分类:AI · News
  • 概况:热度时间:5 hours ago,相关帖子数:4300
  • 是什么事:OpenAI推出面向金融服务行业的ChatGPT，主要服务投资银行家和股票研究员的研究、建模及客户材料制作工作。
  • 为什么重要:这表明生成式AI正从通用助手深入高价值、强监管的金融专业场景，并加剧OpenAI与Anthropic争夺企业级客户的竞争。
  • 讨论概况:X上的讨论主要聚焦于AI是否会取代初级投行人员、金融数据与模型结果的准确性及可追溯性、监管和保密风险，以及OpenAI与Anthropic在金融企业市场的竞争。

话题 4:OpenAI Pauses New $200 ChatGPT Pro Sign-Ups Amid Astra Demand Surge 链接到标题

  • 分类:AI · News
  • 概况:热度时间:1 day ago,相关帖子数:9500
  • 是什么事:OpenAI因代号“Astra”的新功能需求激增，暂停了每月200美元的ChatGPT Pro新用户注册。
  • 为什么重要:这反映出高端AI订阅服务的需求增长可能超过算力和服务容量，也凸显了前沿AI产品商业化、资源调度与用户规模控制之间的矛盾。
  • 讨论概况:X上的讨论主要集中在“Astra”具体能力及其吸引力、暂停注册是算力不足还是营销策略、200美元定价是否合理，以及OpenAI能否稳定提供高端服务。

话题 5:Job Interview Meme Envisions Robot Gunfights in Five Years 链接到标题

  • 分类:AI · Entertainment
  • 概况:热度时间:1 day ago,相关帖子数:15000
  • 是什么事:一则关于“五年后机器人可能发生枪战”的求职面试梗图在 X 平台引发广泛传播。
  • 为什么重要:该话题以夸张的娱乐化方式折射出公众对机器人技术快速发展、人工智能军事化及未来安全风险的关注。
  • 讨论概况:讨论主要围绕这是否只是黑色幽默、机器人和自主武器是否会加速进入现实，以及人工智能发展应如何受到监管展开，支持者与质疑者对其现实可能性存在分歧。

今日 X 上的 AI 舆情小结 链接到标题

今天的舆论主线是，AI正同时沿着“更低成本、更高效率”和“深入企业及专业场景”两条路径加速落地，从模型服务、数据分析到金融研究，商业化竞争明显升温。舆论普遍认可AI应用范围扩大、企业生产效率提升以及高端订阅需求增长的趋势，但对DeepSeek V4.1 Flash的实际性能、OpenAI Data Agent的分析准确性，以及金融场景中的替代效果仍缺乏充分验证。分歧主要集中在高价产品暂停注册究竟是算力供给不足还是营销策略、AI会在多大程度上取代初级专业人员，以及机器人军事化风险究竟是现实威胁还是夸张式表达。潜在风险则包括数据权限与隐私泄露、金融决策错误和责任追溯困难、服务稳定性不足，以及自主系统和AI军事化扩散带来的安全与监管压力。

💡 大佬观点(Influencer Insights) 链接到标题

2026.09.10 AI 行业日度简报 链接到标题

1. 核心趋势：从 Vibe Coding 到 “模型即产品” 的深水区 链接到标题

今日大佬们的讨论焦点已从单纯的模型刷榜，转向了 AI 编程范式的根本性变革与模型商业化的深层博弈。

  • Vibe Coding 进入“工具化”和“长尾化”阶段：@dotey (宝玉) 通过微博 VibeLab 的观察指出，软件开发的门槛已从“专业技能”降维至“语言表达”。核心趋势不再是替代程序员，而是让普通人为满足自身长尾需求而“搓工具”。他特别提到 Shopify 抛弃 React Native 回归原生的案例，揭示了 AI 如何打破技术栈边界，赋能开发者的生产力重构。
  • 模型能力的“消费级落地”与“降智疑云”：@Pluvio9yte 和 @vista8 不约而同地反映了 GPT-6 Astra 在 Codex 中的状态波动。尽管模型上限极高，但“高度降智”和 Token 额度显示异常的现象频发。这引发了关于云端模型稳定性以及服务商是否通过可变内核（如动态降低推理精度）控制成本的猜测。
  • “模型即支付标的”的流量逻辑：@gefei55 (哥飞) 提出了一个尖锐暴论——用户不是为 AI SaaS 付费，而是为底层的 AI 模型付费。结合他分享的 ChatGPT 给产品引流案例，表明在 SEO 与出海领域，绑定顶级模型能力并善于利用 AI 搜索优化，是获取高净值客户的关键。

2. 独特观点与行业前瞻 链接到标题

  • 端侧 AI 与“主动式记忆”的临界点：@zhixianio (知县) 分享了他正在做的面向个人 AI 场景的“舔狗模块”。这标志着个人 AI 正从被动的问答机，进化为通过端侧模型集群实时理解用户环境并主动投喂信息的智能体。他认为消费级设备运行多模型调度是下一个关键战场。
  • 模型竞争的“红皇后效应”：@lijigang (李继刚) 提出模型厂商陷入“落后即死”的红皇后效应博弈。即使市场感觉疲惫，全行业也被迫在极快的迭代速度下奔跑，任何刹车都可能意味着出局。
  • AI 重构交互与设计哲学：@nishuang (倪爽) 和 @lijigang 都提及了交互设计的重心转移。倪爽认为 AI 时代的设计应摒弃复杂的游戏化奖励，回归“游戏感”，让用户因乐趣而非完成任务去消费内容。李继刚则强调，交互应因用户的舞步而移动。

3. 今日推荐的工具与资源 链接到标题

硬核部署与开发

  • Obsidian AI 插件套装：@vista8 利用 GPT-6 Astra 开发了 RSS 订阅、AI 伴读（全格式）及通用 AI 助手插件，适合知识管理用户关注。
  • Flatrouter：@Pluvio9yte 推荐的模型路由平台，主打极致低价（GPT Image 2.5 仅 0.05 元/张），适合做生图应用或批量测试的用户。
  • OpenConnector (开放密码网关)：@ruanyf 推荐的开源工具，解决 AI Agent 密码泄露痛点，统一管理链接授权而不暴露凭据。

内容与效率

  • GPT-Image 2.5 资源汇总：@Pluvio9yte 整理了包括 Sketch 手绘转图、多轮精修、透明背景生成在内的 25+ 条实测爆贴，是目前关于新图像模型最全的一手体验库。
  • 豆包桌面工作套件：@Pluvio9yte 和 @vista8 均在推荐。对于国内非技术背景的普通用户，豆包 PC 版和输入法的成熟度已足以应对日常 AI 介入需求，且没有网络门槛。
  • 美区通讯与支付：@AI_Jasonyu 发布了《2026 美区 PayPal 企业账户开通指南》与 T-Mobile ENC 卡激活教程，这对于完全体验海外全线 AI 服务（尤其是需实体号码验证的层级）是基础建设。

观点与思考： 除了工具，建议关注 @dotey 关于《图解 Skill》的推荐以及 @zhixianio 关于“面向个人 AI 的主动式内存系统”的播客，这是从事 AI 应用层开发需要补齐的底层思维。

📚 附录:今日 Watch List 更新源列表 链接到标题

时间窗口:最近 3 天;覆盖 22 个源;共 38 条更新

Stratechery by Ben Thompson (A_full) 链接到标题

  • The iPhone Duo, The Intelligent Personal Hub, Apple Watch Audio Intelligence
    • 发布时间:2026-09-10 18:00 北京时间
    • 摘要:【待翻译】- Apple once again demonstrated the power of integrating hardware and software, but it’s biggest AI blindspot might be its belief in the primacy of apps.
      • $15 / month or $150 / year.
      • Substantial analysis of the news of the day delivered via three weekly emails or podcasts.
      • Stratechery Interviews.
      • Interviews with leading public CEOs, private company founders, and discussions with fellow analysts.
    • EN 要点:
      • Apple once again demonstrated the power of integrating hardware and software, but it’s biggest AI blindspot might be its belief in the primacy of apps.

OpenAI Blog (A_full) 链接到标题

  • How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules

    • 发布时间:2026-09-11 00:00 北京时间
    • 摘要:【待翻译】- Drug-resistant microbes including bacteria, fungi, parasites, and viruses are a growing global threat.
      • It can take years to find molecules with the potential to become antimicrobials.
      • Researchers are using AI to accelerate this early stage of discovery.
      • “And yet, we haven’t had a new class of antibiotics for 50 years.”.
      • Much of modern antimicrobial development focuses on modifying existing medicines or searching familiar classes of chemicals.
    • EN 要点:
      • César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.

  • Now everyone can put data to work

    • 发布时间:2026-09-10 23:00 北京时间
    • 摘要:【待翻译】- People across every business have questions that data can answer.
      • Where is spending rising?
      • Which issues threaten renewals in our largest accounts, and what should we fix first?
      • Getting those answers often means waiting for a report or asking someone else to run the analysis.
      • It connects to your company data, investigates what changed, and builds interactive dashboards you can share.
    • EN 要点:
      • Meet the Data agent in ChatGPT Work
      • Connect company data, uncover insights, and build interactive dashboards with AI using natural language.

  • Introducing ChatGPT for Financial Services

    • 发布时间:2026-09-10 15:00 北京时间
    • 摘要:【待翻译】- Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.
      • This piece from OpenAI Blog explains how Introducing ChatGPT for Financial Services shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Introducing ChatGPT for Financial Services.
    • EN 要点:
      • Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.

  • Expanding AI access and cyber defense for federal, state, local, and tribal governments

    • 发布时间:2026-09-10 15:00 北京时间
    • 摘要:【待翻译】- OpenAI and GSA will offer eligible federal, state, local, and tribal governments $0 license fees, 50% off usage, and expanded cyber defense support.
      • This piece from OpenAI Blog explains how Expanding AI access and cyber defense for federal, state, local, and tribal governments shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Expanding AI access and cyber defense for federal, state, local, and tribal governments.
    • EN 要点:
      • OpenAI and GSA will offer eligible federal, state, local, and tribal governments $0 license fees, 50% off usage, and expanded cyber defense support.

  • Introducing the Agents API

    • 发布时间:2026-09-10 08:00 北京时间
    • 摘要:【待翻译】- Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.
      • This piece from OpenAI Blog explains how Introducing the Agents API shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Introducing the Agents API.
    • EN 要点:
      • Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.

  • Build more natural voice experiences with GPT‑Live‑1 in the API

    • 发布时间:2026-09-10 08:00 北京时间
    • 摘要:【待翻译】- GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.
      • This piece from OpenAI Blog explains how Build more natural voice experiences with GPT‑Live‑1 in the API shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Build more natural voice experiences with GPT‑Live‑1 in the API.
    • EN 要点:
      • GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.

Two Minute Papers (B_intro+search) 链接到标题

  • I Never Thought I’d See This Happen
    • 发布时间:2026-09-10 16:47 北京时间
    • 摘要:【待翻译】- ❤️ Check out Lambda here and sign up for their GPU Cloud:.
      • 📝 The Navier-Stokes solution paper is available here:.
      • My fluid simulations and papers:.
      • The flow from simulation to reality:.
    • EN 要点:
      • ❤️ Check out Lambda here and sign up for their GPU Cloud:
      • 📝 The Navier-Stokes solution paper is available here:
      • My fluid simulations and papers:
      • The flow from simulation to reality:

ArXiv cs.AI (B_intro+search) 链接到标题

  • OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09203v1 Announce Type: new.
      • Abstract: Existing benchmarks for autonomous AI scientists evaluate only final outputs—generated code, hypotheses, or papers—yet discard the reasoning process by which those outputs were obtained.
      • This makes it impossible to audit scientific methodology, diagnose failure modes, or distinguish systematic reasoning from fortunate guessing.
      • We present \textbf{OpenDiscoveryTrace}, a public dataset of 558 complete AI scientific agent trajectories that captures how models reason, not just what they produce.
    • EN 要点:
      • arXiv:2609.09203v1 Announce Type: new
      • Abstract: Existing benchmarks for autonomous AI scientists evaluate only final outputs—generated code, hypotheses, or papers—yet discard the reasoning proce…
      • This makes it impossible to audit scientific methodology, diagnose failure modes, or distinguish systematic reasoning from fortunate guessing
      • We present \textbf{OpenDiscoveryTrace}, a public dataset of 558 complete AI scientific agent trajectories that captures how models reason, not just what they pr…

  • Adaptive Entangled Game Modules in Artificial General Intelligence

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09226v1 Announce Type: new.
      • Abstract: We introduce a probability-wave framework for modeling the collective behavior of interacting adaptive agents, deriving testable eigenmodes through a generalized behavioral intelligence (GBI) nonlocal probability-wave equation.
      • This framework captures a broad range of human intelligence behaviors with analytical mechanisms and offers an indirect method to examine the Liu-Chen-Ao (LCA) hypothesis of nonlocal entangled nerve fibers in the brain through collective trader behaviors.
      • Our empirical analysis of Chinese intraday stock market data demonstrates that adaptive entangled game modes explain 82-94% (89% overall) of observed decision patterns, a sharp contrast to the predictions of neoclassical finance based on independent rational agents.
    • EN 要点:
      • arXiv:2609.09226v1 Announce Type: new
      • Abstract: We introduce a probability-wave framework for modeling the collective behavior of interacting adaptive agents, deriving testable eigenmodes through a…
      • This framework captures a broad range of human intelligence behaviors with analytical mechanisms and offers an indirect method to examine the Liu-Chen-Ao (LCA)…
      • Our empirical analysis of Chinese intraday stock market data demonstrates that adaptive entangled game modes explain 82-94% (89% overall) of observed decision p…

  • Subagents vs Agent Skills: Executing Reusable Knowledge for Long-Horizon Agentic Tasks

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09233v1 Announce Type: new.
      • Abstract: How can language model agents effectively leverage libraries of reusable knowledge to solve long-horizon tasks?
      • Recent work has increasingly focused on agent skills: reusable capabilities represented as skill packages, i.e., multi-file bundles containing instructions, scripts, and other resources that help agents perform specific tasks.
      • Agent skills are typically executed by loading their skill instructions into an agent’s context and relying on the agent to follow them.
    • EN 要点:
      • arXiv:2609.09233v1 Announce Type: new
      • Abstract: How can language model agents effectively leverage libraries of reusable knowledge to solve long-horizon tasks
      • Recent work has increasingly focused on agent skills: reusable capabilities represented as skill packages, i.e., multi-file bundles containing instructions, scr…
      • Agent skills are typically executed by loading their skill instructions into an agent’s context and relying on the agent to follow them

  • Gradland: On Phenomenal Experience, Differentiated Across Many Dimensions

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09306v1 Announce Type: new.
      • Abstract: This paper investigates the hypothesis that the first-order structure of physical interactions, i.e.
      • gradients or Jacobians, characterizes the structure of phenomenal experience.
      • It does so in an idealized world inhabited by neural networks, Gradland, where the physics are known and the functions are (mostly) differentiable.
    • EN 要点:
      • arXiv:2609.09306v1 Announce Type: new
      • Abstract: This paper investigates the hypothesis that the first-order structure of physical interactions, i.e
      • gradients or Jacobians, characterizes the structure of phenomenal experience
      • It does so in an idealized world inhabited by neural networks, Gradland, where the physics are known and the functions are (mostly) differentiable

  • An Autonomous GeoAI Agent for Arctic Eco-Navigation

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09374v1 Announce Type: new.
      • Abstract: Arctic maritime navigation is becoming increasingly important as changing sea-ice conditions expand seasonal accessibility while simultaneously introducing substantial operational, environmental, and community risks.
      • Arctic route planning is inherently a multi-criteria problem: routes that improve vessel safety or efficiency may increase exposure to sea ice, sensitive ecosystems, or nearby communities.
      • Existing routing methods prioritize travel time, fuel use, and navigational risk, often overlooking ecological and community impacts.
    • EN 要点:
      • arXiv:2609.09374v1 Announce Type: new
      • Abstract: Arctic maritime navigation is becoming increasingly important as changing sea-ice conditions expand seasonal accessibility while simultaneously introd…
      • Arctic route planning is inherently a multi-criteria problem: routes that improve vessel safety or efficiency may increase exposure to sea ice, sensitive ecosys…
      • Existing routing methods prioritize travel time, fuel use, and navigational risk, often overlooking ecological and community impacts

  • The Menu Is an Execution Prior: State-Path Tool Menus for Online Agents

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09395v1 Announce Type: new.
      • Abstract: Language models act through tools, yet practical agents face libraries containing thousands of interfaces.
      • We introduce the tool menu as the short, ordered subset of available tools shown to an agent before execution.
      • The agent can call only tools in this menu.
    • EN 要点:
      • arXiv:2609.09395v1 Announce Type: new
      • Abstract: Language models act through tools, yet practical agents face libraries containing thousands of interfaces
      • We introduce the tool menu as the short, ordered subset of available tools shown to an agent before execution
      • The agent can call only tools in this menu

  • Decision-Focused Active Learning for Scale-Aware Critical-Materials Recovery

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09413v1 Announce Type: new.
      • Abstract: Choosing a recovery process for scale-up requires connecting laboratory results with product requirements, process costs, and scale effects.
      • We analyze records from Pacific Northwest National Laboratory’s Computer Intelligence for Critical Element Recovery and Optimization (CICERO) workflow for autonomous selective precipitation.
      • Active learning uses prior results to choose experiments.
    • EN 要点:
      • arXiv:2609.09413v1 Announce Type: new
      • Abstract: Choosing a recovery process for scale-up requires connecting laboratory results with product requirements, process costs, and scale effects
      • We analyze records from Pacific Northwest National Laboratory’s Computer Intelligence for Critical Element Recovery and Optimization (CICERO) workflow for auton…
      • Active learning uses prior results to choose experiments

  • Valerant: An Automatic Navigable Game Map Generator via Action-Conditioned World Model Exploration

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09418v1 Announce Type: new.
      • Abstract: World Action Models (WAMs) couple predictive world modeling with action generation, allowing anticipated future states to guide agent behavior.
      • Although WAMs are rapidly advancing embodied AI, general-purpose counterparts remain largely unexplored in games.
      • Existing game-oriented approaches often combine action-conditioned world models with external policies and reward functions to realize WAM-like decision-making, yet they operate mainly in 2D visual observation space and do not instantiate persistent 3D geometry.
    • EN 要点:
      • arXiv:2609.09418v1 Announce Type: new
      • Abstract: World Action Models (WAMs) couple predictive world modeling with action generation, allowing anticipated future states to guide agent behavior
      • Although WAMs are rapidly advancing embodied AI, general-purpose counterparts remain largely unexplored in games
      • Existing game-oriented approaches often combine action-conditioned world models with external policies and reward functions to realize WAM-like decision-making,…

  • XAI-Arena: Can LLMs Assess the Quality of XAI Explanations?

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09428v1 Announce Type: new.
      • Abstract: Evaluating the quality of explanations produced by explainable AI (XAI) methods remains challenging because existing approaches often rely on subjective human judgment, limiting reproducibility, scalability, and comparability between studies.
      • We examine whether LLMs can serve as a reproducible and scalable mechanism to make comparative assessments of the quality of XAI explanations.
      • We introduce XAI-Arena, an LLM-as-a-judge framework for scalable, reproducible, multidimensional, and stakeholder-sensitive evaluation of XAI explanation quality.
    • EN 要点:
      • arXiv:2609.09428v1 Announce Type: new
      • Abstract: Evaluating the quality of explanations produced by explainable AI (XAI) methods remains challenging because existing approaches often rely on subjecti…
      • We examine whether LLMs can serve as a reproducible and scalable mechanism to make comparative assessments of the quality of XAI explanations
      • We introduce XAI-Arena, an LLM-as-a-judge framework for scalable, reproducible, multidimensional, and stakeholder-sensitive evaluation of XAI explanation qualit…

  • Do Agents Know When They Succeed? Calibrating Agent Confidence from Internal Representations

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09448v1 Announce Type: new.
      • Abstract: As agentic systems getting adopted rapidly in safety critical applications, it is vital to measure the confidence associated with the agentic actions.
      • In comparison to the traditional machine learning systems, agentic workflows have complex failure modes with planning, tool invocation and dynamic environment interactions.
      • In this paper, we investigate whether model’s internal representations provide stronger signals of eventual task success in multi-turn agentic setups.
    • EN 要点:
      • arXiv:2609.09448v1 Announce Type: new
      • Abstract: As agentic systems getting adopted rapidly in safety critical applications, it is vital to measure the confidence associated with the agentic actions
      • In comparison to the traditional machine learning systems, agentic workflows have complex failure modes with planning, tool invocation and dynamic environment i…
      • In this paper, we investigate whether model’s internal representations provide stronger signals of eventual task success in multi-turn agentic setups

ArXiv cs.CL (B_intro+search) 链接到标题

  • X-CoSD: Communication-Efficient Cross-Vocabulary Collaborative Speculative Decoding

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09166v1 Announce Type: new.
      • Abstract: This paper investigates collaborative speculative decoding (CoSD), a distributed large language model (LLM) inference framework in which an on-device small language model (SLM) drafts candidate tokens and a server LLM verifies them.
      • Existing CoSD methods assume a shared vocabulary between the SLM and the LLM and incur substantial communication load because residual resampling requires token distribution exchange between the user device and the edge server.
      • To address these limitations, we propose cross-vocabulary CoSD (X-CoSD), a lossless and communication-efficient CoSD framework for heterogeneous SLM-LLM vocabularies.
    • EN 要点:
      • arXiv:2609.09166v1 Announce Type: new
      • Abstract: This paper investigates collaborative speculative decoding (CoSD), a distributed large language model (LLM) inference framework in which an on-device…
      • Existing CoSD methods assume a shared vocabulary between the SLM and the LLM and incur substantial communication load because residual resampling requires token…
      • To address these limitations, we propose cross-vocabulary CoSD (X-CoSD), a lossless and communication-efficient CoSD framework for heterogeneous SLM-LLM vocabul…

  • StochBench: A Domain-Specific Benchmark for Stochastic Processes in Lean

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09264v1 Announce Type: new.
      • Abstract: Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications.
      • We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language source.
      • Addressing a field underrepresented in Mathlib, it covers finite and countable Markov chains, renewal processes, random walks, martingales, stopping times, queues, Brownian motion, stochastic calculus, weak convergence, and Poisson and continuous-time Markov processes.
    • EN 要点:
      • arXiv:2609.09264v1 Announce Type: new
      • Abstract: Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam…
      • We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language s…
      • Addressing a field underrepresented in Mathlib, it covers finite and countable Markov chains, renewal processes, random walks, martingales, stopping times, queu…

  • Osprey: Target-agnostic Pre-training Makes Stronger Drafters in Speculative Decoding

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09338v1 Announce Type: new.
      • Abstract: Speculative decoding is critical for accelerating LLM inference.
      • However, the speedup is fragile: drafters are typically trained against a narrow distribution for a single target model, and their acceptance rate collapses under workload shifts.
      • This is a striking inversion of modern LLM development, where target models are valued precisely for the broad generalization they acquire through large-scale pretraining.
    • EN 要点:
      • arXiv:2609.09338v1 Announce Type: new
      • Abstract: Speculative decoding is critical for accelerating LLM inference
      • However, the speedup is fragile: drafters are typically trained against a narrow distribution for a single target model, and their acceptance rate collapses und…
      • This is a striking inversion of modern LLM development, where target models are valued precisely for the broad generalization they acquire through large-scale p…

  • SWORD: Wikidata-based Distortions Reveal Hidden Cross-Lingual Inconsistencies in LLM Factual Error Rejection

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09349v1 Announce Type: new.
      • Abstract: Modern LLMs demonstrate impressive multilingual performance, yet standard benchmarks primarily reward selecting correct answers rather than evaluating genuine factual understanding.
      • We introduce Systematic Wikidata-based Object-Relation Distortion (SWORD), a benchmark that evaluates whether models consistently reject factual errors across languages.
      • SWORD generates syntactically well-formed but factually incorrect statements in eight widely spoken languages through controlled perturbations of Wikidata triples, ranging from random entity substitutions to semantically plausible property-based selections.
    • EN 要点:
      • arXiv:2609.09349v1 Announce Type: new
      • Abstract: Modern LLMs demonstrate impressive multilingual performance, yet standard benchmarks primarily reward selecting correct answers rather than evaluating…
      • We introduce Systematic Wikidata-based Object-Relation Distortion (SWORD), a benchmark that evaluates whether models consistently reject factual errors across l…
      • SWORD generates syntactically well-formed but factually incorrect statements in eight widely spoken languages through controlled perturbations of Wikidata tripl…

  • Auditable Emergency Triage for Maternal and Newborn Care in India

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09356v1 Announce Type: new.
      • Abstract: At Noora Health, our nurses answer more than 50,000 medical queries per month on our WhatsApp-based service that provides caregivers with on-demand support.
      • Their most time-critical task is emergency triage: deciding which queries need immediate in-person attention.
      • To support them, we built a system that uses a large language model (LLM) to classify whether a message is an emergency and provide a rationale for interpretability.
    • EN 要点:
      • arXiv:2609.09356v1 Announce Type: new
      • Abstract: At Noora Health, our nurses answer more than 50,000 medical queries per month on our WhatsApp-based service that provides caregivers with on-demand su…
      • Their most time-critical task is emergency triage: deciding which queries need immediate in-person attention
      • To support them, we built a system that uses a large language model (LLM) to classify whether a message is an emergency and provide a rationale for interpretabi…

  • Do LLMs Make More Mistakes If They Do Not Believe the Input Data?

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09363v1 Announce Type: new.
      • Abstract: Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or data-to-text systems.
      • We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict).
      • To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on local knowledge, only partially captured in models’ parametric knowledge.
    • EN 要点:
      • arXiv:2609.09363v1 Announce Type: new
      • Abstract: Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or d…
      • We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict)
      • To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on loca…

  • Benchmarking Hybrid Deep Research Across Database Querying and Web Search

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09410v1 Announce Type: new.
      • Abstract: While autonomous agents have made significant strides in “deep research” by iteratively navigating the open web to synthesize information, real-world problem-solving is rarely confined to a single environment.
      • Complex analytical tasks inherently require agents to weave together evidence from both ambiguous unstructured text (e.g., the open web) and highly precise structured data (e.g., relational databases).
      • However, existing benchmarks evaluate these modalities in isolation, failing to capture the critical “handoff” - the ability to preserve constraints when moving evidence between systems.
    • EN 要点:
      • arXiv:2609.09410v1 Announce Type: new
      • Abstract: While autonomous agents have made significant strides in “deep research” by iteratively navigating the open web to synthesize information, real-world…
      • Complex analytical tasks inherently require agents to weave together evidence from both ambiguous unstructured text (e.g., the open web) and highly precise stru…
      • However, existing benchmarks evaluate these modalities in isolation, failing to capture the critical “handoff” - the ability to preserve constraints when moving…

  • Edu-QuRating: Multi-Dimensional Educational Data Curation with Distilled Pairwise Judgements

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09425v1 Announce Type: new.
      • Abstract: Educational data filters have become a practical way to improve language-model pre-training, but most filters treat educational value as a single scalar property.
      • This may be too broad for some applications, especially if the data set already features a high density of educational material.
      • Useful learning material needs to be accurate, engaging, well structured, and appropriate for the intended audience and application (e.g.
    • EN 要点:
      • arXiv:2609.09425v1 Announce Type: new
      • Abstract: Educational data filters have become a practical way to improve language-model pre-training, but most filters treat educational value as a single scal…
      • This may be too broad for some applications, especially if the data set already features a high density of educational material
      • Useful learning material needs to be accurate, engaging, well structured, and appropriate for the intended audience and application (e.g

  • The Mutations of Machine Speech

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09496v1 Announce Type: new.
      • Abstract: Algorithmic outputs now populate the digital environments through which contemporary life is organized.
      • The role of law in facilitating and constituting (rather than merely responding to) these processes is gaining increasing traction across scholarly accounts.
      • This inquiry traces the evolution of algorithmic outputs attending to their legal underpinnings and social implications, surfacing the mutations of machine speech.
    • EN 要点:
      • arXiv:2609.09496v1 Announce Type: new
      • Abstract: Algorithmic outputs now populate the digital environments through which contemporary life is organized
      • The role of law in facilitating and constituting (rather than merely responding to) these processes is gaining increasing traction across scholarly accounts
      • This inquiry traces the evolution of algorithmic outputs attending to their legal underpinnings and social implications, surfacing the mutations of machine spee…

  • TEFM: Token-Efficient Faithful Modeling for Structured Data

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09552v1 Announce Type: new.
      • Abstract: In this paper, we solve two fundamental obstacles in applying LLMs to critical domains: token efficiency and faithfulness.
      • To address both constraints jointly, we present TEFM (Token-Efficient Faithful Modeling), a framework designed for structured data analysis in critical domains.
      • TEFM achieves token efficiency by compressing lengthy structured observations into compact Behavioral Code tokens, dramatically reducing token consumption with minimal information loss.
    • EN 要点:
      • arXiv:2609.09552v1 Announce Type: new
      • Abstract: In this paper, we solve two fundamental obstacles in applying LLMs to critical domains: token efficiency and faithfulness
      • To address both constraints jointly, we present TEFM (Token-Efficient Faithful Modeling), a framework designed for structured data analysis in critical domains
      • TEFM achieves token efficiency by compressing lengthy structured observations into compact Behavioral Code tokens, dramatically reducing token consumption with…

ArXiv cs.LG (B_intro+search) 链接到标题

  • Spectral origin of the topological gap exponent d + {\eta}: mechanism, kernel, decomposition, and scope

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09159v1 Announce Type: new.
      • Abstract: The topological gap $\Delta$ – the excess $H_1$ total persistence of a critical point cloud over a density-matched null – scales as $\Delta \sim L^{d+\eta}$.
      • We derive this analytically: the spectral integral $I(\alpha) = \sum_{k\neq 0} S_{\mathrm{conn}}(k),|k|^{\alpha}$ scales as $L^{2-\alpha-\eta}$ when IR-dominated, giving $I(-2\eta) \sim L^{d+\eta}$.
      • The decomposition $I(-2\eta) = I_0 \cdot I_{\mathrm{shape}}$ separates volume ($I_0 \propto N(1-m^2) \sim L^d$) from anomalous dimension ($I_{\mathrm{shape}} \sim L^{\eta}$); the volume factor accounts for the magnetization-driven per-configuration variance of $\Delta$.
    • EN 要点:
      • arXiv:2609.09159v1 Announce Type: new
      • Abstract: The topological gap $\Delta$ – the excess $H_1$ total persistence of a critical point cloud over a density-matched null – scales as $\Delta \sim L^{…
      • We derive this analytically: the spectral integral $I(\alpha) = \sum_{k\neq 0} S_{\mathrm{conn}}(k),|k|^{\alpha}$ scales as $L^{2-\alpha-\eta}$ when IR-dominat…
      • The decomposition $I(-2\eta) = I_0 \cdot I_{\mathrm{shape}}$ separates volume ($I_0 \propto N(1-m^2) \sim L^d$) from anomalous dimension ($I_{\mathrm{shape}} \s…

  • Physics-informed neural networks by Gradient-Guided Gaussian Adaptive Sampling (3GAS-PINNs)

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09162v1 Announce Type: new.
      • Abstract: Physics-informed neural networks (PINNs) provide a mesh-free framework for solving partial differential equations, yet their performance in nonlinear problems is often limited by slow convergence, gradient imbalance, and insufficient resolution to capture localized intermittent structures such as shock waves[1].
      • These issues arise primarily from the use of fixed weights of loss and uniform collocation point distributions, which cannot adapt to the evolving complexity of the solution field during training.
      • To address these challenges, Gradient-Guided Gaussian Adaptive Sampling Physics-Informed Neural Networks (3GAS-PINNs) is proposed in this paper, which combines uniform probability distribution and Gaussian-smoothed probability distribution derived from the spatial gradients of solution, to maintain global constraint satisfaction as well as concentrating collocation points in regions of high gradient.
    • EN 要点:
      • arXiv:2609.09162v1 Announce Type: new
      • Abstract: Physics-informed neural networks (PINNs) provide a mesh-free framework for solving partial differential equations, yet their performance in nonlinear…
      • These issues arise primarily from the use of fixed weights of loss and uniform collocation point distributions, which cannot adapt to the evolving complexity of…
      • To address these challenges, Gradient-Guided Gaussian Adaptive Sampling Physics-Informed Neural Networks (3GAS-PINNs) is proposed in this paper, which combines…

  • World-Time Compute with Verified Code World Models

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09163v1 Announce Type: new.
      • Abstract: LLMs generalize across a domain only after seeing many real, labeled examples, which most domains lack.
      • We study a way to manufacture it cheaply.
      • When a domain’s dynamics can be written as code, one template instantiates into many world models: executable, verifiable programs over symbolic state, each an inexhaustible source of exactly-labeled trajectories.
    • EN 要点:
      • arXiv:2609.09163v1 Announce Type: new
      • Abstract: LLMs generalize across a domain only after seeing many real, labeled examples, which most domains lack
      • We study a way to manufacture it cheaply
      • When a domain’s dynamics can be written as code, one template instantiates into many world models: executable, verifiable programs over symbolic state, each an…

  • Scaling Post-Training Ternarisation to Qwen3-8B Capability Retention, Reproduction, Lossless Packing, and Packed Execution

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09240v1 Announce Type: new.
      • Abstract: Ultra-low-bit language models promise reductions in storage and memory traffic, but a nominal “1.58-bit” label does not specify the deployed representation or its execution cost.
      • We study a scale-up of an aggressive post-training conversion pipeline from Qwen3-4B to Qwen3-8B.
      • The conversion uses KOTMS rotation, E2M-ATQ adaptive ternarisation, and GPTQ-style error compensation in a weight-only A16 configuration.
    • EN 要点:
      • arXiv:2609.09240v1 Announce Type: new
      • Abstract: Ultra-low-bit language models promise reductions in storage and memory traffic, but a nominal “1.58-bit” label does not specify the deployed represent…
      • We study a scale-up of an aggressive post-training conversion pipeline from Qwen3-4B to Qwen3-8B
      • The conversion uses KOTMS rotation, E2M-ATQ adaptive ternarisation, and GPTQ-style error compensation in a weight-only A16 configuration

  • Distribution-Consistent Inference for Dynamic Sparse Mixture-of-Experts

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09241v1 Announce Type: new.
      • Abstract: Mixture-of-Experts (MoE) architectures have emerged as a powerful paradigm for scaling model capacity while preserving efficient inference in large foundation models.
      • However, most MoE models use a fixed top-$k$ expert selection policy, assigning the same expert budget to every token even when fewer experts may be sufficient.
      • Inference-time dynamic top-$k$ routing can reduce computation without retraining, but existing methods often overlook the distributional shift caused by deviating from the training-time routing configuration.
    • EN 要点:
      • arXiv:2609.09241v1 Announce Type: new
      • Abstract: Mixture-of-Experts (MoE) architectures have emerged as a powerful paradigm for scaling model capacity while preserving efficient inference in large fo…
      • However, most MoE models use a fixed top-$k$ expert selection policy, assigning the same expert budget to every token even when fewer experts may be sufficient
      • Inference-time dynamic top-$k$ routing can reduce computation without retraining, but existing methods often overlook the distributional shift caused by deviati…

  • DiffLUT-Net: Differentiable Training of FPGA LUT Networks with Learnable Connectivity

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09254v1 Announce Type: new.
      • Abstract: Field-programmable gate arrays (FPGAs) enable efficient neural-network inference, but most deployment flows either accelerate multiply-accumulate operations or convert pretrained quantized models into lookup tables (LUTs).
      • We present DiffLUT-Net, an FPGA-native network connected by six-input LUTs that are trained from scratch.
      • We jointly learn the 64 truth-table entries of a LUT and the source to each of its six input ports using a differentiable LUT function relaxation and hardware source selection.
    • EN 要点:
      • arXiv:2609.09254v1 Announce Type: new
      • Abstract: Field-programmable gate arrays (FPGAs) enable efficient neural-network inference, but most deployment flows either accelerate multiply-accumulate oper…
      • We present DiffLUT-Net, an FPGA-native network connected by six-input LUTs that are trained from scratch
      • We jointly learn the 64 truth-table entries of a LUT and the source to each of its six input ports using a differentiable LUT function relaxation and hardware s…

  • Accountable and uncertainty-aware evaluation of sensor-based AI under distribution shift: devices, subjects, and nearly three years underground

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09257v1 Announce Type: new.
      • Abstract: Sensor-based AI systems are rarely operated under the conditions under which they were trained: devices, personnel and recording epochs change, and each change degrades performance in ways a random train-test split cannot reveal.
      • We propose a staged, accountable evaluation protocol that treats the evaluation of a deployed model as a measurement with declared reference levels and a quantified uncertainty.
      • Four cumulative generalisation stages hold out devices, subjects and time.
    • EN 要点:
      • arXiv:2609.09257v1 Announce Type: new
      • Abstract: Sensor-based AI systems are rarely operated under the conditions under which they were trained: devices, personnel and recording epochs change, and ea…
      • We propose a staged, accountable evaluation protocol that treats the evaluation of a deployed model as a measurement with declared reference levels and a quanti…
      • Four cumulative generalisation stages hold out devices, subjects and time

  • Literati: Towards Anytime Optimal Shape Generalized Trees via AO*

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09299v1 Announce Type: new.
      • Abstract: Decision trees are prized for their interpretability and strong performance on tabular data, but popular greedy top-down induction algorithms can yield suboptimal and unnecessarily complex structures.
      • Optimal decision tree methods address this through global optimization, yet remain restricted to axis-aligned threshold splits, which limit the expressivity of each node and often force deep, complex trees to capture non-linear feature effects.
      • Shape Generalized Trees (SGTs) generalize threshold splits to learnable univariate shape functions, improving expressivity and enabling more compact trees.
    • EN 要点:
      • arXiv:2609.09299v1 Announce Type: new
      • Abstract: Decision trees are prized for their interpretability and strong performance on tabular data, but popular greedy top-down induction algorithms can yiel…
      • Optimal decision tree methods address this through global optimization, yet remain restricted to axis-aligned threshold splits, which limit the expressivity of…
      • Shape Generalized Trees (SGTs) generalize threshold splits to learnable univariate shape functions, improving expressivity and enabling more compact trees

  • Explaining f-Divergence-Based Regularization via Local Curvature and Sharpness-Aware Minimization

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09367v1 Announce Type: new.
      • Abstract: Divergence-based regularization and Sharpness-Aware Minimization (SAM) are two prominent approaches for improving generalization in deep learning, both motivated by robustness to perturbations.
      • However, their relationship has remained largely unexplored.
      • Building on classical second-order expansions of $f$-divergences, we show that the two methods are locally consistent under parameter-space perturbations: both induce curvature-sensitive penalties, with divergence regularization yielding a Fisher-weighted quadratic form and SAM penalizing sharpness through the dominant Hessian eigenvalue.
    • EN 要点:
      • arXiv:2609.09367v1 Announce Type: new
      • Abstract: Divergence-based regularization and Sharpness-Aware Minimization (SAM) are two prominent approaches for improving generalization in deep learning, bot…
      • However, their relationship has remained largely unexplored
      • Building on classical second-order expansions of $f$-divergences, we show that the two methods are locally consistent under parameter-space perturbations: both…

  • Constraint-Aware Discrete Black-Box Optimization Using Tensor Decomposition

    • 发布时间:2026-09-10 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.09370v1 Announce Type: new.
      • Abstract: Discrete black-box optimization is often addressed using approaches such as Sequential Model-Based Optimization (SMBO), which aims to improve sample efficiency by fitting surrogate models that approximate a costly objective function over a discrete search space.
      • In many real-world problems, the set of feasible inputs is often given by logical constraints known in advance.
      • However, existing surrogate modeling techniques generally fail to capture the symbolic rules governing feasibility in discrete input spaces.
    • EN 要点:
      • arXiv:2609.09370v1 Announce Type: new
      • Abstract: Discrete black-box optimization is often addressed using approaches such as Sequential Model-Based Optimization (SMBO), which aims to improve sample e…
      • In many real-world problems, the set of feasible inputs is often given by logical constraints known in advance
      • However, existing surrogate modeling techniques generally fail to capture the symbolic rules governing feasibility in discrete input spaces