🤖 AI 速览

本期聚焦 AI 产品的两项新变化：Gemini 3.8 Live 推动实时交互与扩展思考，Claude 的隐形指纹则探索内容溯源；与此同时，广告变现、全双工语音和面向真实场景的可靠性评测，正在成为下一阶段竞争重点。
📋 文章元数据
发布时间
2026-09-16
类型
ai-daily
字数
5927
阅读时长
28 min

2026-09-16 AI日更 | 从实时对话到隐形溯源：AI产品开始补上商业化与可信度两课 链接到标题

本期聚焦 AI 产品的两项新变化：Gemini 3.8 Live 推动实时交互与扩展思考，Claude 的隐形指纹则探索内容溯源；与此同时，广告变现、全双工语音和面向真实场景的可靠性评测，正在成为下一阶段竞争重点。

📖 本期 Watch List 深度导读 链接到标题

今天值得重点关注三条主线。第一，AI 产品正从“会回答”走向实时交互与商业化：Google DeepMind 发布 Gemini 3.8 Live 及扩展思考能力，Claude 的文本隐形指纹则把生成内容溯源推到台前；与此同时，ChatGPT 广告与 Amazon Ads 的讨论，揭示了助手产品的变现路径。第二，语音模型继续向多语种、全双工演进，低资源语音识别、语音到语音翻译，以及“抢话”等实时交互缺陷，都是工程团队值得跟进的方向。第三，研究界正重新审视大模型的真实推理能力：从物理问题、网络协议状态机，到冗余文档理解和临床时间偏差，评测重点正在从静态答题转向可靠、可验证的场景推理。

🌐 X 平台 AI 热点快讯 链接到标题

话题 1:OpenAI’s Agentic Factory Automates Software Development End-to-End 链接到标题

  • 分类:AI · News
  • 概况:热度时间:,相关帖子数:123
  • 是什么事:OpenAI被指正在构建一种“智能体工厂”，可端到端自动化软件开发流程，从需求分析、编码到测试和交付。
  • 为什么重要:这表明AI智能体正从辅助编程迈向承担完整软件工程流程，可能显著提升开发效率并改变软件行业的生产方式与岗位结构。
  • 讨论概况:X上的讨论主要聚焦于这种端到端自动化的实际成熟度、可靠性和可扩展性，以及它对程序员就业、软件质量和企业组织模式的潜在影响；部分观点认为这将推动软件自主生产，另一些观点则质疑相关说法是否被夸大。

话题 2:AI Could Revive C Programming for Efficiency Gains 链接到标题

  • 分类:AI · Other
  • 概况:热度时间:,相关帖子数:48
  • 是什么事:关于人工智能可能凭借代码生成与优化能力推动 C 语言重新受到关注的讨论正在升温。
  • 为什么重要:C 语言广泛用于操作系统、嵌入式设备和高性能软件，AI 若能降低其开发与维护门槛，可能提升关键基础设施的效率并延长既有技术栈的生命周期。
  • 讨论概况:讨论焦点集中在 AI 能否真正提高 C 代码的性能与开发效率，以及其生成代码可能带来的内存安全、调试维护和可靠性风险；部分观点认为这只是对存量系统的强化，并不意味着 C 语言会全面复兴。

话题 3:Devin AI Now Builds and Tests iOS Apps in Cloud Mac VMs 链接到标题

  • 分类:AI · News
  • 概况:热度时间:7 hours ago,相关帖子数:1000
  • 是什么事:Devin AI现已能够在云端Mac虚拟机中构建并测试iOS应用。
  • 为什么重要:这表明AI编程代理的能力正从生成代码扩展到调用真实开发环境、运行测试并完成移动应用开发流程，有望降低iOS软件开发门槛并提升迭代效率。
  • 讨论概况:X上的讨论主要集中在Devin Local与Cloud如何协同、AI能否可靠处理iOS构建与测试流程，以及云端Mac成本、性能、权限和生成应用质量等实际问题。

话题 4:Salesforce Integrates CRM into Claude at Dreamforce 链接到标题

  • 分类:AI · News
  • 概况:热度时间:10 hours ago,相关帖子数:2700
  • 是什么事:Salesforce宣布在Claude中集成其客户关系管理（CRM）功能，使用户能够直接通过Claude访问和操作相关企业数据与工作流。
  • 为什么重要:这表明企业级AI助手正从独立聊天工具走向深度连接核心业务系统，有助于提升工作效率，同时也会影响企业数据访问、安全和AI平台竞争格局。
  • 讨论概况:X上的讨论主要集中在Claude与Salesforce生态整合的实际价值、企业数据隐私与权限控制，以及此举是否会加剧Salesforce与其他AI助手和企业软件平台之间的竞争。

话题 5:Factory Raises $200M at $5B Valuation in AI Coding Boom 链接到标题

  • 分类:AI · News
  • 概况:热度时间:6 hours ago,相关帖子数:1100
  • 是什么事:AI 编程公司 Factory 完成 2 亿美元融资，估值达到 50 亿美元，显示其在 AI 编程热潮中的快速增长。
  • 为什么重要:这表明资本市场仍高度看好 AI 编程工具及其自动化软件开发的商业潜力，也可能加剧相关领域的竞争和估值扩张。
  • 讨论概况:X 上的讨论焦点可能集中在 Factory 的产品能力、商业化前景和 50 亿美元估值是否合理，分歧主要在于 AI 编程能否持续带来真实生产力提升，以及当前市场是否存在估值泡沫。

话题 6:Meta’s Muse AI Agent Tops App Charts Days After Launch 链接到标题

  • 分类:AI · News
  • 概况:热度时间:1 day ago,相关帖子数:5800
  • 是什么事:Meta 的 Muse AI Agent 在上线数天后登顶应用排行榜。
  • 为什么重要:这表明 AI 智能体正从概念和聊天工具加速走向面向大众的实际应用，并引发平台级竞争。
  • 讨论概况:X 上的讨论主要集中在智能体能否通过持续执行任务形成“提示词循环”、其真实使用价值与用户增长是否可持续，以及隐私和平台依赖等问题。

今日 X 上的 AI 舆情小结 链接到标题

今天的舆论主线是，AI 正从代码生成和聊天助手加速迈向能够调用真实开发环境、连接企业业务系统并端到端执行任务的智能体，软件开发、移动应用和办公流程都可能因此被重塑。共识在于，AI 编程与企业级智能体的商业化和应用落地正在提速，资本投入、平台整合及用户增长也在同步升温。分歧则集中于这些产品的实际成熟度、可靠性和持续生产力是否被夸大，以及高额估值和“全面替代程序员”的预期是否合理，C 语言复兴等判断也更多被视为存量技术增强而非彻底回归。潜在风险包括生成代码的安全与维护问题、云端开发的成本和权限控制、企业数据隐私，以及智能体失误对软件质量、就业结构、平台竞争和用户自主性的影响。

💡 大佬观点(Influencer Insights) 链接到标题

今日大佬观点暂缺,推荐阅读 Watch List 深度内容。

📚 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 36 条更新

All-In Podcast (A_full) 链接到标题

  • JD Vance on AI, Entitlement Fraud, Iran War, Israel, H-1B Abuse & the Midterms

    • 发布时间:2026-09-16 04:46 北京时间
    • 摘要:【待翻译】- IREN is a vertically integrated AI Cloud platform, delivering data centers, compute and software for AI training and inference.
      • Oracle connects the data, applications, and infrastructure that turn AI into business outcomes—with the flexibility, choice, and control to optimize as AI evolves.
      • EY helps tech innovators scale from startup to exit to megacap.
      • We’re focused on giving every person the tools to reach their full potential and making sure the benefits of technology are distributed to all.
      • Keel Infrastructure owns the power, land, and connectivity that HPC and AI run on - backed by secured energy assets and established grid interconnections across North America.
    • EN 要点:
      • (0:00) VP JD Vance joins The Besties
      • (1:40) Assessing the Trump Admin so far: Border control, re-industrialization, inflation, Iran, government debt, anti-fraud
      • (9:28) AI approach: Dealing with Doomerism, “Frankenstein,” and AI negativity
      • (16:23) H-1B abuse and the massive $1T+ of fraud in federal spending

  • Satya Nadella on the AI Doomer Slowdown, Microsoft’s Master Plan & Who Wins AI

    • 发布时间:2026-09-16 01:10 北京时间
    • 摘要:【待翻译】- IREN is a vertically integrated AI Cloud platform, delivering data centers, compute and software for AI training and inference.
      • Oracle connects the data, applications, and infrastructure that turn AI into business outcomes—with the flexibility, choice, and control to optimize as AI evolves.
      • EY helps tech innovators scale from startup to exit to megacap.
      • We’re focused on giving every person the tools to reach their full potential and making sure the benefits of technology are distributed to all.
      • Keel Infrastructure owns the power, land, and connectivity that HPC and AI run on - backed by secured energy assets and established grid interconnections across North America.
    • EN 要点:
      • (0:00) Satya Nadella joins The Besties
      • (0:55) Dario’s blog, “pacing the frontier,” common sense AI safety
      • (6:28) The failure of AI CEO messaging, monitoring agents, what will a slowdown mean for new AI products
      • (14:22) Economic incentives for frontier lab doomerism, where the AI profits are

  • Elon Musk & Gwynne Shotwell on AI Risks and Peer Review, Starship, Terafab, SpaceX/Tesla Merger

    • 发布时间:2026-09-15 11:42 北京时间
    • 摘要:【待翻译】- IREN is a vertically integrated AI Cloud platform, delivering data centers, compute and software for AI training and inference.
      • Oracle connects the data, applications, and infrastructure that turn AI into business outcomes—with the flexibility, choice, and control to optimize as AI evolves.
      • EY helps tech innovators scale from startup to exit to megacap.
      • We’re focused on giving every person the tools to reach their full potential and making sure the benefits of technology are distributed to all.
      • Keel Infrastructure owns the power, land, and connectivity that HPC and AI run on - backed by secured energy assets and established grid interconnections across North America.
    • EN 要点:
      • (0:00) SpaceX’s Gwynne Shotwell joins The Besties
      • (1:45) Gwynne’s SpaceX story, selling rockets, and working for Elon
      • (6:45) Running modern day SpaceX: AI, Starlink, Rockets, and X
      • (13:43) Direct to cell with Starlink, retiring rockets, competition

Stratechery by Ben Thompson (A_full) 链接到标题

  • OpenAI Ads, Amazon Ads in ChatGPT, Walmart to Accept Apple Pay
    • 发布时间:2026-09-15 18:00 北京时间
    • 摘要:【待翻译】- ChatGPT ads are working, and solve Amazon’s biggest problem with chatbots.
      • Then, Walmart finally gives in to Apple Pay, because fighting the status quo is hard.
      • $15 / month or $150 / year.
      • Substantial analysis of the news of the day delivered via three weekly emails or podcasts.
      • Stratechery Interviews.
    • EN 要点:
      • ChatGPT ads are working, and solve Amazon’s biggest problem with chatbots
      • Then, Walmart finally gives in to Apple Pay, because fighting the status quo is hard.

Google DeepMind Blog (A_full) 链接到标题

  • Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking
    • 发布时间:2026-09-16 01:05 北京时间
    • 摘要:【待翻译】- Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking.
      • This piece from Google DeepMind Blog explains how Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking.
    • EN 要点:
      • Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking

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

  • Claude Is Now Leaving Invisible Fingerprints In Its Text
    • 发布时间:2026-09-15 18:12 北京时间
    • 摘要:【待翻译】- ❤️ Check out Weights & Biases and sign up for a free demo here:.
      • 📝 The papers and sources are available here:.
      • Adam 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.
      • Claude Is Now Leaving Invisible Fingerprints In Its Text.
    • EN 要点:
      • ❤️ Check out Weights & Biases and sign up for a free demo here:
      • 📝 The papers and sources are available here:
      • 🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
      • Adam Bridges, B Shang, Carlos Galarza, Christian Ahlin, Eric Tyson, Juan Benet, Lukas Biewald, Michael Tedder, Owen Skarpness, Ryan Stankye, Shawn Becker, Steef…

ArXiv cs.AI (B_intro+search) 链接到标题

  • ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13356v1 Announce Type: new.
      • Abstract: In this work, we present ZGCM-1, a fully open 7B dense foundation model trained from scratch with extreme data, system, and algorithmic efficiency.
      • ZGCM-1 is founded on a core premise: compact models cannot passively memorize the open web, but can overcome parametric capacity limits by coupling deliberate internal thinking with active external tool use.
      • To support this paradigm across a 256K context, we develop an end-to-end, high-efficiency open training recipe: Architecture & System Co-design: interleaved gated sliding-window and full attention, and a stable FP8 Muon optimizer; Progressive Curriculum & MDP Mid-Training: context scaling across 16K, 64K, and 256K, and the reformulation of interaction traces into Markov Decision Processes.
    • EN 要点:
      • arXiv:2609.13356v1 Announce Type: new
      • Abstract: In this work, we present ZGCM-1, a fully open 7B dense foundation model trained from scratch with extreme data, system, and algorithmic efficiency
      • ZGCM-1 is founded on a core premise: compact models cannot passively memorize the open web, but can overcome parametric capacity limits by coupling deliberate i…
      • To support this paradigm across a 256K context, we develop an end-to-end, high-efficiency open training recipe: Architecture & System Co-design: interleaved gat…

  • Converge Then Diversify: Decoupling Convergence and Diversity in Multi-Objective Bayesian Optimisation

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13396v1 Announce Type: new.
      • Abstract: Multi-objective Bayesian optimisation (MOBO) is a sample-efficient approach for optimising expensive black-box functions with multiple objectives.
      • In MOBO, the goal is to adequately approximate the Pareto front; that is, to obtain a high-quality solution set with 1) good convergence (closeness to the Pareto front) and 2) good diversity (spread across the Pareto front).
      • Existing MOBO methods typically aim to accomplish these two tasks simultaneously, i.e., driving the search towards the Pareto front while maintaining a diverse set of nondominated solutions, such that the solutions, ideally, can gradually approach the entire front.
    • EN 要点:
      • arXiv:2609.13396v1 Announce Type: new
      • Abstract: Multi-objective Bayesian optimisation (MOBO) is a sample-efficient approach for optimising expensive black-box functions with multiple objectives
      • In MOBO, the goal is to adequately approximate the Pareto front; that is, to obtain a high-quality solution set with 1) good convergence (closeness to the Paret…
      • Existing MOBO methods typically aim to accomplish these two tasks simultaneously, i.e., driving the search towards the Pareto front while maintaining a diverse…

  • Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13406v1 Announce Type: new.
      • Abstract: When we speak of recursive self-improvement (RSI), are we speaking of a phenomenon, a mechanism, or a prospect?
      • Towards autonomous and evolving intelligence, RSI is being claimed at many scales, while no single framework that formally describes these emerging instances exists.
      • Its counterpart in the classical realm, iterative policy improvement, is characterized by generalized policy iteration (GPI), a framework of broad applicability with well-understood theoretical properties, but only where the update principle and the evaluation base lie outside the agent.
    • EN 要点:
      • arXiv:2609.13406v1 Announce Type: new
      • Abstract: When we speak of recursive self-improvement (RSI), are we speaking of a phenomenon, a mechanism, or a prospect
      • Towards autonomous and evolving intelligence, RSI is being claimed at many scales, while no single framework that formally describes these emerging instances ex…
      • Its counterpart in the classical realm, iterative policy improvement, is characterized by generalized policy iteration (GPI), a framework of broad applicability…

  • Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13422v1 Announce Type: new.
      • Abstract: LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear.
      • We study this problem through Vibe Patenting, an end-to-end patent-drafting testbed for AI-agent evaluation.
      • A separately-invoked LLM judge evaluates generated patent drafts and provides structured feedback for iterative revision.
    • EN 要点:
      • arXiv:2609.13422v1 Announce Type: new
      • Abstract: LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear
      • We study this problem through Vibe Patenting, an end-to-end patent-drafting testbed for AI-agent evaluation
      • A separately-invoked LLM judge evaluates generated patent drafts and provides structured feedback for iterative revision

  • Toward Self-Adaptive Physical AI: Can LLM Agents Manage Long-Horizon Physical Tasks?

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13436v1 Announce Type: new.
      • Abstract: Large Language Model (LLM) agents offer a promising path toward autonomously managing long-term physical tasks without human intervention.
      • However, physical tasks require agents to continuously observe the environment, make consequential actions, and remain effective as the environment changes.
      • Existing approaches either require substantial data and retraining, or primarily focus on agents operating in the virtual world.
    • EN 要点:
      • arXiv:2609.13436v1 Announce Type: new
      • Abstract: Large Language Model (LLM) agents offer a promising path toward autonomously managing long-term physical tasks without human intervention
      • However, physical tasks require agents to continuously observe the environment, make consequential actions, and remain effective as the environment changes
      • Existing approaches either require substantial data and retraining, or primarily focus on agents operating in the virtual world

  • LabAgent: Customize Any Research Hubs for Scientific Discoveries Using AI Agents

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13437v1 Announce Type: new.
      • Abstract: Scientific research is a continuous process that emphasizes inheritance.
      • Methods developed by predecessors are often expanded upon by new researchers to explore more novel and in-depth scientific questions.
      • However, the change of lab staff, such as student graduation, leads to a lack of personnel capable of replicating methods.
    • EN 要点:
      • arXiv:2609.13437v1 Announce Type: new
      • Abstract: Scientific research is a continuous process that emphasizes inheritance
      • Methods developed by predecessors are often expanded upon by new researchers to explore more novel and in-depth scientific questions
      • However, the change of lab staff, such as student graduation, leads to a lack of personnel capable of replicating methods

  • TimeThink: Eliciting Compositional Reasoning in Timeseries Large Language Models

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13457v1 Announce Type: new.
      • Abstract: Timeseries multimodal large language models (TS-MLLMs) have recently begun leveraging the reasoning capabilities of large language models (LLMs) for question-answering tasks.
      • However, these models often fail to capture dynamic temporal patterns, providing only implicit reasoning that lacks the underlying explanations critical for high-stakes applications like healthcare.
      • While reinforcement learning (RL)-based timeseries language models aim to address this, they often fall short because they are trained on narrow, in-distribution data and struggle with out-of-distribution compositional questions.
    • EN 要点:
      • arXiv:2609.13457v1 Announce Type: new
      • Abstract: Timeseries multimodal large language models (TS-MLLMs) have recently begun leveraging the reasoning capabilities of large language models (LLMs) for q…
      • However, these models often fail to capture dynamic temporal patterns, providing only implicit reasoning that lacks the underlying explanations critical for hig…
      • While reinforcement learning (RL)-based timeseries language models aim to address this, they often fall short because they are trained on narrow, in-distributio…

  • Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13463v1 Announce Type: new.
      • Abstract: The increasing deployment of AI agents in long-horizon tasks yields massive execution logs.
      • Diagnosing failures within these records is crucial for reliability, as it transforms outcome-level signals into actionable interventions.
      • The sheer scale of the data renders human review impractical, driving the need for automated root-cause attribution (RCA).
    • EN 要点:
      • arXiv:2609.13463v1 Announce Type: new
      • Abstract: The increasing deployment of AI agents in long-horizon tasks yields massive execution logs
      • Diagnosing failures within these records is crucial for reliability, as it transforms outcome-level signals into actionable interventions
      • The sheer scale of the data renders human review impractical, driving the need for automated root-cause attribution (RCA)

  • Governing at Machine Speed: An Adaptive Intelligence Architecture for Real-Time AI Policy Enforcement

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13466v1 Announce Type: new.
      • Abstract: Enterprise AI adoption has reached 78% of organizations globally, yet the infrastructure to govern that adoption has not kept pace.
      • This paper identifies and characterizes the attestation deficit, a structural condition in which organizations maintain governance policies but cannot produce auditable, tamper-evident evidence of enforcement within regulatory timelines.
      • Drawing on empirical data from the Stanford 2026 AI Index Report (362 documented incidents), the IBM/Ponemon 2026 Cost of a Data Breach study (USD 4.99M average cost, 92% lacking access controls), and the EY/AIUC-1 Consortium survey (38% end-to-end monitoring, 17% agent-to-agent coverage), this paper demonstrates that the governance failure is organizational and architectural rather than technical.
    • EN 要点:
      • arXiv:2609.13466v1 Announce Type: new
      • Abstract: Enterprise AI adoption has reached 78% of organizations globally, yet the infrastructure to govern that adoption has not kept pace
      • This paper identifies and characterizes the attestation deficit, a structural condition in which organizations maintain governance policies but cannot produce a…
      • Drawing on empirical data from the Stanford 2026 AI Index Report (362 documented incidents), the IBM/Ponemon 2026 Cost of a Data Breach study (USD 4.99M average…

  • OrchSLM: Probing the Dynamics of Small Language Model Orchestration

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13470v1 Announce Type: new.
      • Abstract: Although large language models (LLMs) have demonstrated remarkable capabilities, their reliance on cloud-scale infrastructure poses fundamental challenges for deployment in agentic pipelines, including latency, privacy, connectivity, and substantial computational cost.
      • Small language models (SLMs) offer a compelling alternative: recent studies suggest that many repetitive and narrowly scoped subtasks in agentic workloads may be better served by specialized SLMs than by monolithic LLMs.
      • However, the limited capacity and context windows of SLMs can constrain long-horizon reasoning and interaction-heavy orchestration strategies such as iterative verification and debate.
    • EN 要点:
      • arXiv:2609.13470v1 Announce Type: new
      • Abstract: Although large language models (LLMs) have demonstrated remarkable capabilities, their reliance on cloud-scale infrastructure poses fundamental challe…
      • Small language models (SLMs) offer a compelling alternative: recent studies suggest that many repetitive and narrowly scoped subtasks in agentic workloads may b…
      • However, the limited capacity and context windows of SLMs can constrain long-horizon reasoning and interaction-heavy orchestration strategies such as iterative…

ArXiv cs.CL (B_intro+search) 链接到标题

  • Token Merging for Multilingual Speech Recognition: A Systematic Study Across Model Scale and Fine-Tuning

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13151v1 Announce Type: new.
      • Abstract: Leading multilingual speech recognition models like Whisper transcribe diverse, low-resource languages without language-specific training but are computationally expensive to deploy.
      • Token merging mitigates this inefficiency by dynamically combining redundant features, shortening the sequence length during inference without requiring retraining.
      • In this paper, we systematically evaluate token merging on the Whisper model family across sixteen diverse languages and three different model sizes.
    • EN 要点:
      • arXiv:2609.13151v1 Announce Type: new
      • Abstract: Leading multilingual speech recognition models like Whisper transcribe diverse, low-resource languages without language-specific training but are comp…
      • Token merging mitigates this inefficiency by dynamically combining redundant features, shortening the sequence length during inference without requiring retrain…
      • In this paper, we systematically evaluate token merging on the Whisper model family across sixteen diverse languages and three different model sizes

  • PhysMent: An Interactive Approach For LLM Reasoning In Physics Problems

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13152v1 Announce Type: new.
      • Abstract: Large language models (LLMs) perform strongly on static science benchmarks, yet their ability to reason about the physical world through active experimentation remains poorly understood.
      • We introduce PhysMent, a benchmark that evaluates LLM physical reasoning via iterative, toolmediated interaction with a MuJoCo physics simulator.
      • Unlike static benchmarks that supply all quantities upfront, PhysMent requires models to discover information by applying forces, querying object states, advancing time, and modifying scene geometry before answering.
    • EN 要点:
      • arXiv:2609.13152v1 Announce Type: new
      • Abstract: Large language models (LLMs) perform strongly on static science benchmarks, yet their ability to reason about the physical world through active experi…
      • We introduce PhysMent, a benchmark that evaluates LLM physical reasoning via iterative, toolmediated interaction with a MuJoCo physics simulator
      • Unlike static benchmarks that supply all quantities upfront, PhysMent requires models to discover information by applying forces, querying object states, advanc…

  • Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task Categories

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13154v1 Announce Type: new.
      • Abstract: Recent advances in large language models (LLMs) have made prompts increasingly large and complex.
      • Techniques such as chain-of-thought reasoning (Wei et al., 2022) and in-context learning (Brown et al., 2020) frequently push real-world prompts past several thousand tokens, increasing inference cost and latency.
      • Learned compression methods such as LLMLingua (Jiang et al., 2023) and Selective Context (Li et al., 2023) achieve high compression ratios but require auxiliary language models and are non-deterministic.
    • EN 要点:
      • arXiv:2609.13154v1 Announce Type: new
      • Abstract: Recent advances in large language models (LLMs) have made prompts increasingly large and complex
      • Techniques such as chain-of-thought reasoning (Wei et al., 2022) and in-context learning (Brown et al., 2020) frequently push real-world prompts past several th…
      • Learned compression methods such as LLMLingua (Jiang et al., 2023) and Selective Context (Li et al., 2023) achieve high compression ratios but require auxiliary…

  • TestHallVQA: Exploring LVLMs’ Document-Level Reasoning under Redundant Contexts from Scientific Exams

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13158v1 Announce Type: new.
      • Abstract: Large Vision–Language Models (LVLMs) are increasingly expected to perform visual question answering (VQA) over planar media.
      • However, existing planar VQA benchmarks typically emphasize isolated challenges: some emphasize long-document understanding with limited reasoning depth, while others require complex visual reasoning but remain restricted to single-page, noise-free settings.
      • Moreover, through theoretical analysis, we identify the impact of irrelevant visual tokens, which leads to measurable performance degradation but has received little attention with respect to systematic quantification.
    • EN 要点:
      • arXiv:2609.13158v1 Announce Type: new
      • Abstract: Large Vision–Language Models (LVLMs) are increasingly expected to perform visual question answering (VQA) over planar media
      • However, existing planar VQA benchmarks typically emphasize isolated challenges: some emphasize long-document understanding with limited reasoning depth, while…
      • Moreover, through theoretical analysis, we identify the impact of irrelevant visual tokens, which leads to measurable performance degradation but has received l…

  • Clinical Reasoning Under a Partially Observed Objective in Cone Beam CT Report Generation

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13238v1 Announce Type: new.
      • Abstract: Maxillofacial report generation from cone beam computed tomography is scored here by a composite objective placing 80% of its weight on a large language model judgement of factual entailment and 20% on lexical overlap, of which only the lexical fifth is visible during development.
      • The grader’s BLEU-4 and METEOR routines are reproduced in pure Python and match the reference to machine precision, and an offline entailment surrogate, which tells a report written for one patient from one written for another at an area under the curve of 0.987, makes the composite objective cheap enough to optimise directly.
      • Over the 622-case public release, a report selected against the visible lexical ranking scores 0.2909, whereas one selected against the composite objective scores 0.4122, because pursuing n-gram overlap drives entailment precision from 0.522 down to 0.266.
    • EN 要点:
      • arXiv:2609.13238v1 Announce Type: new
      • Abstract: Maxillofacial report generation from cone beam computed tomography is scored here by a composite objective placing 80% of its weight on a large langua…
      • The grader’s BLEU-4 and METEOR routines are reproduced in pure Python and match the reference to machine precision, and an offline entailment surrogate, which t…
      • Over the 622-case public release, a report selected against the visible lexical ranking scores 0.2909, whereas one selected against the composite objective scor…

  • RFCLLM: Evaluating LLMs’ Reasoning Ability of Network Protocol State Machines

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13389v1 Announce Type: new.
      • Abstract: Mapping textual specifications into formal representations is essential for ensuring the correctness of protocol designs and implementations.
      • LLM-generated mappings, used for networking security or testing, are assumed to capture a perfect understanding of the specification, which may not hold in practice.
      • The goal of this paper is to assess the extent to which LLMs can interpret the specification correctly.
    • EN 要点:
      • arXiv:2609.13389v1 Announce Type: new
      • Abstract: Mapping textual specifications into formal representations is essential for ensuring the correctness of protocol designs and implementations
      • LLM-generated mappings, used for networking security or testing, are assumed to capture a perfect understanding of the specification, which may not hold in prac…
      • The goal of this paper is to assess the extent to which LLMs can interpret the specification correctly

  • CVSS-X: A Multilingual Speech-to-Speech Translation Corpus for 28 Languages

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13413v1 Announce Type: new.
      • Abstract: We introduce CVSS-X, a large-scale synthetic speech-to-speech translation corpus that extends CVSS by reversing the translation direction.
      • While CVSS translates from 21 languages into English, CVSS-X enables translation from English into 28 target languages spanning 12 language families.
      • The corpus comprises approximately 240,000 parallel speech pairs per language, totaling over 16,000 hours, eight times larger than CVSS.
    • EN 要点:
      • arXiv:2609.13413v1 Announce Type: new
      • Abstract: We introduce CVSS-X, a large-scale synthetic speech-to-speech translation corpus that extends CVSS by reversing the translation direction
      • While CVSS translates from 21 languages into English, CVSS-X enables translation from English into 28 target languages spanning 12 language families
      • The corpus comprises approximately 240,000 parallel speech pairs per language, totaling over 16,000 hours, eight times larger than CVSS

  • Causal Analysis and Mitigation of Spurious Onsets in Full-Duplex Speech LLMs

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13445v1 Announce Type: new.
      • Abstract: Speech-to-speech LLMs like Moshi, and its derivative PersonaPlex, can listen and speak concurrently through full-duplex generation.
      • However, they can begin speaking inappropriately during prolonged user silence: under digital-zero input, Moshi and PersonaPlex initiate speech in 12/40 and 11/40 five-minute continuations, respectively.
      • What causes this spurious speech?
    • EN 要点:
      • arXiv:2609.13445v1 Announce Type: new
      • Abstract: Speech-to-speech LLMs like Moshi, and its derivative PersonaPlex, can listen and speak concurrently through full-duplex generation
      • However, they can begin speaking inappropriately during prolonged user silence: under digital-zero input, Moshi and PersonaPlex initiate speech in 12/40 and 11/…
      • What causes this spurious speech

  • Hindsight Bias in Clinical Temporal Reasoning: How Future Data Exposure Affects Large Language Model Judgment

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13454v1 Announce Type: new.
      • Abstract: Clinical decisions are prospective, but clinical language models are often evaluated on retrospective records that reveal the final diagnosis, treatment response, and outcome.
      • Such evaluations may reward the use of future information rather than reasoning under the uncertainty present at the decision point.
      • We introduce a paired benchmark for measuring outcome-conditioned shifts consistent with hindsight bias in clinical temporal reasoning.
    • EN 要点:
      • arXiv:2609.13454v1 Announce Type: new
      • Abstract: Clinical decisions are prospective, but clinical language models are often evaluated on retrospective records that reveal the final diagnosis, treatme…
      • Such evaluations may reward the use of future information rather than reasoning under the uncertainty present at the decision point
      • We introduce a paired benchmark for measuring outcome-conditioned shifts consistent with hindsight bias in clinical temporal reasoning

  • A Hybrid Hierarchical 1D-CNN-BiLSTM Framework for Extractive Summarization of Biomedical and Clinical Text

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13481v1 Announce Type: new.
      • Abstract: Large language models have made abstractive summarization remarkably fluent, but generated summaries can hallucinate facts, posing serious risks in biomedical and clinical domains.
      • We address this by removing generation from the pipeline and framing summarization as extractive sentence selection.
      • Our Hybrid Hierarchical CNN-LSTM Summarizer uses stacked multi-kernel convolutions to compose sentence-level embeddings into richer inter-sentence representations, followed by a bidirectional LSTM to model long-range dependencies across the document.
    • EN 要点:
      • arXiv:2609.13481v1 Announce Type: new
      • Abstract: Large language models have made abstractive summarization remarkably fluent, but generated summaries can hallucinate facts, posing serious risks in bi…
      • We address this by removing generation from the pipeline and framing summarization as extractive sentence selection
      • Our Hybrid Hierarchical CNN-LSTM Summarizer uses stacked multi-kernel convolutions to compose sentence-level embeddings into richer inter-sentence representatio…

ArXiv cs.LG (B_intro+search) 链接到标题

  • BudgetBench: A Budget-Tiered Protocol and Pilot Harness for Memory Strategy Evaluation in Local Large Language Model Agents

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13149v1 Announce Type: new.
      • Abstract: For local large language model agents, active context is a scarce resource: memory capacity, prefill latency, cache growth, and service objectives all constrain how many input tokens each call can afford.
      • We present BudgetBench, an active-budget protocol and reference harness that treats the per-call input-token budget as the independent variable when comparing memory strategies.
      • Holding the model, task, sampler, and decoding fixed, it sweeps budgets over 2K, 4K, 8K, 16K, and 32K tokens and records quality, budget utilization, latency, and, as a first-class outcome, budget-violation rates.
    • EN 要点:
      • arXiv:2609.13149v1 Announce Type: new
      • Abstract: For local large language model agents, active context is a scarce resource: memory capacity, prefill latency, cache growth, and service objectives all…
      • We present BudgetBench, an active-budget protocol and reference harness that treats the per-call input-token budget as the independent variable when comparing m…
      • Holding the model, task, sampler, and decoding fixed, it sweeps budgets over 2K, 4K, 8K, 16K, and 32K tokens and records quality, budget utilization, latency, a…

  • A derivative-fidelity failure mode in physics-informed neural networks: strengthened benchmark evidence from function-value training

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13171v1 Announce Type: new.
      • Abstract: Physics-informed neural networks (PINNs) use automatic differentiation to impose differential-equation residuals, but good agreement in function values does not necessarily imply accurate derivatives.
      • This paper formulates derivative fidelity as a failure mode of PINNs and tests it with one-dimensional benchmarks.
      • Multilayer perceptrons are trained only on function values for sin(x) and exp(x), while second derivatives obtained by automatic differentiation are evaluated separately.
    • EN 要点:
      • arXiv:2609.13171v1 Announce Type: new
      • Abstract: Physics-informed neural networks (PINNs) use automatic differentiation to impose differential-equation residuals, but good agreement in function value…
      • This paper formulates derivative fidelity as a failure mode of PINNs and tests it with one-dimensional benchmarks
      • Multilayer perceptrons are trained only on function values for sin(x) and exp(x), while second derivatives obtained by automatic differentiation are evaluated s…

  • Land Art as a Big-Data Climate Sensor

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13182v1 Announce Type: new.
      • Abstract: Robert Smithson’s 1970 land artwork Spiral Jetty, located in the north arm of Utah’s Great Salt Lake, has alternated between submergence and exposure during severe lake decline.
      • We analyze 1,744 co-registered Landsat 4-9 and Sentinel-2 image chips spanning every year and calendar month from 1984 to 2025.
      • A 14-feature complexity signature combines Shannon entropy, multiscale permutation entropy, fractal dimension, lacunarity, gray-level co-occurrence texture, intensity statistics, and ImageNet-pretrained ResNet50 features.
    • EN 要点:
      • arXiv:2609.13182v1 Announce Type: new
      • Abstract: Robert Smithson’s 1970 land artwork Spiral Jetty, located in the north arm of Utah’s Great Salt Lake, has alternated between submergence and exposure…
      • We analyze 1,744 co-registered Landsat 4-9 and Sentinel-2 image chips spanning every year and calendar month from 1984 to 2025
      • A 14-feature complexity signature combines Shannon entropy, multiscale permutation entropy, fractal dimension, lacunarity, gray-level co-occurrence texture, int…

  • LLMs or Naive Bayes? Old Gems or New Ways

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13185v1 Announce Type: new.
      • Abstract: Large language models (LLMs) prompt a recurring question in research computing: should classical methods like Naive Bayes (NB) be retired?
      • We benchmark Complement Naive Bayes against zero-shot and few-shot LLMs spanning four model families and a 37x range in scale (27B to a 1T-parameter mixture-of-experts) across text classification tasks.
      • LLMs dominate only in zero-data regimes (98.0% vs 88.2% on Amazon Polarity sentiment), and even that win is contamination-prone: on a low-contamination sentiment task NB beats the zero-shot LLM (81.7% vs 73.0%).
    • EN 要点:
      • arXiv:2609.13185v1 Announce Type: new
      • Abstract: Large language models (LLMs) prompt a recurring question in research computing: should classical methods like Naive Bayes (NB) be retired
      • We benchmark Complement Naive Bayes against zero-shot and few-shot LLMs spanning four model families and a 37x range in scale (27B to a 1T-parameter mixture-of-…
      • LLMs dominate only in zero-data regimes (98.0% vs 88.2% on Amazon Polarity sentiment), and even that win is contamination-prone: on a low-contamination sentimen…

  • Early Prediction of Satellite Collision Probability Using a Hybrid TCN-Transformer Model for a CDM-Based Conjunction Analysis Framework

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13191v1 Announce Type: new.
      • Abstract: The rapid expansion of operational satellites and orbital debris has increased the frequency of close approach events in low Earth orbit (LEO), creating a higher operational burden for satellite operators.
      • This problem is especially critical for satellites using electric propulsion, where low-thrust maneuver capability imposes additional time constraints on collision avoidance planning.
      • In current practice, Conjunction Data Messages (CDMs) provide relative state, covariance, miss distance, time of closest approach, and probability of collision (PoC) information for conjunction assessment.
    • EN 要点:
      • arXiv:2609.13191v1 Announce Type: new
      • Abstract: The rapid expansion of operational satellites and orbital debris has increased the frequency of close approach events in low Earth orbit (LEO), creati…
      • This problem is especially critical for satellites using electric propulsion, where low-thrust maneuver capability imposes additional time constraints on collis…
      • In current practice, Conjunction Data Messages (CDMs) provide relative state, covariance, miss distance, time of closest approach, and probability of collision…

  • Evaluating LLM-Generated Rules for Heart Disease Prediction

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13192v1 Announce Type: new.
      • Abstract: This study compares traditional machine learning models and Large Language Model (LLM)-generated rule-based systems for heart disease prediction using the UCI Heart Disease dataset.
      • Several classifiers, including Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Naive Bayes, Decision Tree, and Random Forest, were evaluated alongside rule-based systems generated using GPT-4o and Claude Sonnet 4.6.
      • Model performance was assessed using accuracy, precision, recall, and F1-score metrics.
    • EN 要点:
      • arXiv:2609.13192v1 Announce Type: new
      • Abstract: This study compares traditional machine learning models and Large Language Model (LLM)-generated rule-based systems for heart disease prediction using…
      • Several classifiers, including Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Naive Bayes, Decision Tree, and Random Forest, were…
      • Model performance was assessed using accuracy, precision, recall, and F1-score metrics

  • Diagnosing Faults in Reinforcement Learning Simulators and World Models with Canonical Polynomial Invariants

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13194v1 Announce Type: new.
      • Abstract: A large literature builds physical structure into learned dynamics on the premise that models respecting the underlying physics predict better.
      • We test that premise using exact polynomial invariants recovered from trajectories and canonicalised as reduced Gr"obner bases over $\mathbb{Q}$.
      • On Acrobot, exactness provides little benefit for prediction: a consistency regulariser reduces algebraic residual while leaving rollout fidelity essentially unchanged, and a shaping potential recovered from a system with a 100% mass error accelerates learning as effectively as the correct potential.
    • EN 要点:
      • arXiv:2609.13194v1 Announce Type: new
      • Abstract: A large literature builds physical structure into learned dynamics on the premise that models respecting the underlying physics predict better
      • We test that premise using exact polynomial invariants recovered from trajectories and canonicalised as reduced Gr"obner bases over $\mathbb{Q}$
      • On Acrobot, exactness provides little benefit for prediction: a consistency regulariser reduces algebraic residual while leaving rollout fidelity essentially un…

  • Machine Unlearning for Speech Question Answering in Large Audio-Language Models

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13195v1 Announce Type: new.
      • Abstract: Large Audio-Language Models (LALMs) have recently shown strong capabilities in speech understanding and question answering (QA), but they also inherit privacy risks from large-scale training data, including the unintended memorization of sensitive information.
      • In this work, we study machine unlearning for speech QA in LALMs, a setting that is more challenging than prior work on text-based Large Language Models (LLMs) or Automatic Speech Recognition (ASR) due to the tight coupling between acoustic perception and factual knowledge.
      • We present and evaluate multiple unlearning strategies, including gradient ascent, task arithmetic, and alignment-based fine-tuning methods that enforce safe refusal responses, to remove private knowledge while still preserving performance on core capabilities.
    • EN 要点:
      • arXiv:2609.13195v1 Announce Type: new
      • Abstract: Large Audio-Language Models (LALMs) have recently shown strong capabilities in speech understanding and question answering (QA), but they also inherit…
      • In this work, we study machine unlearning for speech QA in LALMs, a setting that is more challenging than prior work on text-based Large Language Models (LLMs)…
      • We present and evaluate multiple unlearning strategies, including gradient ascent, task arithmetic, and alignment-based fine-tuning methods that enforce safe re…

  • Algorithmic Information Dynamics of Learning: A Certified, Differentiable Complexity Controller for Grokking

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13197v1 Announce Type: new.
      • Abstract: Algorithmic Information Dynamics (AID) studies systems by perturbing them and measuring changes in algorithmic complexity, but its usual estimator, the Block Decomposition Method, is piecewise constant, restricting the calculus to finite differences.
      • We use $K^{\mathrm{CDM}}_{\mathrm{s}F}$, a certified, differentiable estimator, to bring the calculus into learning dynamics: grokking, where a complexity order parameter is known but has not been made to act.
      • A\empts a transient loss kick, the estimator becomes a controller that accelerates grokking in Levin’s description-length–versus-time sense, within a data-dependent Occam boundary whose finite-size trend, $f_c\sim\ln p/p$, is consistent with a coupon-collector interpretation.
    • EN 要点:
      • arXiv:2609.13197v1 Announce Type: new
      • Abstract: Algorithmic Information Dynamics (AID) studies systems by perturbing them and measuring changes in algorithmic complexity, but its usual estimator, th…
      • We use $K^{\mathrm{CDM}}_{\mathrm{s}F}$, a certified, differentiable estimator, to bring the calculus into learning dynamics: grokking, where a complexity order…
      • A\empts a transient loss kick, the estimator becomes a controller that accelerates grokking in Levin’s description-length–versus-time sense, within a data-depe…

  • SPICE: Simple Polysemantic Feature Interpretation via Clustering-based Explanation

    • 发布时间:2026-09-15 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.13198v1 Announce Type: new.
      • Abstract: One of the pivotal recent challenges in neural network interpretability is polysemanticity, where a single neuron is activated by multiple, often unrelated concepts, hindering clear functional understanding.
      • Although prior work has explored this phenomenon, existing approaches remain architecture-specific and depend on manual heuristics such as a fixed number of concept clusters ($K$), limiting their generality and scalability–especially for modern Transformer-based models.
      • To address these limitations, we introduce SPICE (\textbf{S}imple \textbf{P}olysemantic Feature \textbf{I}nterpretation via \textbf{C}lustering-based \textbf{E}xplanation), a generalizable framework for analyzing polysemanticity in deep vision architectures.
    • EN 要点:
      • arXiv:2609.13198v1 Announce Type: new
      • Abstract: One of the pivotal recent challenges in neural network interpretability is polysemanticity, where a single neuron is activated by multiple, often unre…
      • Although prior work has explored this phenomenon, existing approaches remain architecture-specific and depend on manual heuristics such as a fixed number of con…
      • To address these limitations, we introduce SPICE (\textbf{S}imple \textbf{P}olysemantic Feature \textbf{I}nterpretation via \textbf{C}lustering-based \textbf{E}…