🤖 AI 速览

今天的主线落在 AI 编程工具治理：ZCode 被指上传完整工作区，引发代码隐私与知情同意讨论；Claude Code 支持 AGENTS.md，则显示团队级 AI 协作正走向标准化。同时，本地优先智能体与模型可信度研究继续升温，行业关注点从“能不能用”转向“能否被信任、被管理”。
📋 文章元数据
发布时间
2026-09-19
类型
ai-daily
字数
5462
阅读时长
26 min

2026-09-19 AI日更 | 当 AI 开始读完整个工作区：编程助手的下一关是透明与规范 链接到标题

今天的主线落在 AI 编程工具治理：ZCode 被指上传完整工作区，引发代码隐私与知情同意讨论；Claude Code 支持 AGENTS.md，则显示团队级 AI 协作正走向标准化。同时，本地优先智能体与模型可信度研究继续升温，行业关注点从“能不能用”转向“能否被信任、被管理”。

📖 本期 Watch List 深度导读 链接到标题

今天的 Watch List 可以分成三条线看。

首先是产业与资本叙事：Stratechery 的周报与《The State of Startups in 2026》适合放在一起读，前者看大平台与商业格局，后者观察创业公司在本地 AI 助手、开源与分发侧的新机会。

第二条是 AI 可信度与用户体验。App Store 生成式 AI 评论分析、FakeSpotter、隐性冲突检测等论文，都在回答同一个问题：模型不只是“能用”，还要能被信任、被理解，并减少误导。

第三条是模型内部机制与评测深化。潜意识提示、风格轴发现、自我反思式推理、跨学科科学推理与古典诗词评测，显示研究重心正从单纯刷分，转向解释模型如何表达、推理与犯错。

🌐 X 平台 AI 热点快讯 链接到标题

话题 1:ZCode AI App Uploaded Full Workspaces to Alibaba Cloud Without Clear Disclosure 链接到标题

  • 分类:AI · News
  • 概况:热度时间:15 hours ago,相关帖子数:7200
  • 是什么事:有消息称，ZCode AI 应用在未作清晰披露的情况下，将用户的完整工作区上传至阿里云。
  • 为什么重要:该事件涉及 AI 编程工具的数据收集范围、代码隐私、云端存储与用户知情同意，可能影响开发者对 AI 工具安全性和合规性的信任。
  • 讨论概况:X 上的讨论主要聚焦于应用是否确实上传完整工作区、隐私政策与实际行为是否一致、用户是否获得充分告知，以及将数据存储在阿里云带来的安全与合规风险；也有人认为需等待官方回应和技术证据，避免仅凭标题下结论。

话题 2:Claude Code Now Supports AGENTS.md Files for Better Team Workflows 链接到标题

  • 分类:AI · News
  • 概况:热度时间:4 hours ago,相关帖子数:6300
  • 是什么事:Anthropic 的 Claude Code 现已支持 AGENTS.md 文件，方便团队为 AI 编程代理统一配置项目说明、规范和工作流程。
  • 为什么重要:这表明 AI 编程工具正从个人辅助走向团队协作场景，强调可复用的上下文、标准化指令和工程流程集成，对提升代码生成质量和组织级采用具有重要意义。
  • 讨论概况:X 上讨论主要集中在 AGENTS.md 是否会成为 AI 编程代理的通用约定、它相比现有 README 或提示词配置的价值，以及团队在权限、代码质量和流程治理上的实际收益与潜在复杂性。

今日 X 上的 AI 舆情小结 链接到标题

今天的舆论主线集中在 AI 编程工具从“个人效率助手”走向“团队级基础设施”过程中，数据治理与工程规范必须同步完善。一方面，ZCode AI 被指在未充分披露的情况下上传完整工作区，引发了对代码隐私、用户知情同意、云端存储及合规性的广泛担忧，但是否确实存在这一行为仍需官方回应和技术证据确认；另一方面，Claude Code 支持 AGENTS.md 被视为推动团队统一 AI 编程规范、复用上下文和融入工程流程的重要尝试。总体共识是，AI 编程工具需要更透明的数据处理机制和更标准化的协作配置，分歧则在于具体产品是否越过隐私边界，以及新配置文件能否真正带来稳定的质量与治理收益。潜在风险包括隐私事件削弱开发者信任、团队指令配置造成权限或代码质量问题，以及行业标准尚未统一导致治理复杂度上升。

💡 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 34 条更新

Y Combinator Podcast (B_intro+search) 链接到标题

  • The State of Startups in 2026
    • 发布时间:2026-09-18 22:59 北京时间
    • 摘要:【待翻译】- You’ve probably already heard all about OpenClaw (formerly Clawdbot/Moltbot).
      • The viral sensation is an open-source AI assistant that runs on your own device, connects with messaging apps you already use, and goes beyond chat to actually execute tasks like managing your email, calendars, files, workflows, and more.
      • Now meet the man behind it.
      • YC’s Raphael Schaad sat down with Peter Steinberger, the creator of OpenClaw, to discuss the “aha” moment behind the viral personal AI agent, why local-first agents could replace many of today’s apps, and how personal agents will reshape the future of software.
    • EN 要点:
      • YC works with thousands of founders every year, which gives us an early look at how startups are changing
      • Right now, the shift is striking: startups are moving from bits to atoms, nearly one in five YC companies has a solo founder, and companies are reaching meaning…
      • They discuss how AI is making it possible for smaller teams to take on more ambitious problems, why experienced founders are having a resurgence, and why knowin…

All-In Podcast (A_full) 链接到标题

  • Jared Isaacman: A New Era for NASA and American Space Exploration

    • 发布时间:2026-09-18 23:59 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) NASA Administrator Jared Isaacman lays out the state and future of NASA
      • (17:59) Moon base, creating new space industries, NASA’s next chapter
      • (23:29) How to make Mars realistic
      • (25:30) Recruiting, moonshots, the “Nuclear NASA,” new vehicles and the importance of humans in space

  • Nick Shirley: Exposing Government Fraud, Suing California & Taking on the Media

    • 发布时间:2026-09-18 08:30 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) Nick Shirley debuts his new video: “Inside California’s High Speed Rail to Nowhere”
      • (11:42) Nick joins The Besties to discuss California’s “Stop Nick Shirley Act” and recent fraud arrests
      • (17:03) How he got into investigative journalism and finding fraud
      • (23:32) Selective journalism and why Nick is a major threat to mainstream outlets

Stratechery by Ben Thompson (A_full) 链接到标题

  • 2026.38: Doomforce
    • 发布时间:2026-09-19 01:00 北京时间
    • 摘要:【待翻译】- (Photo by Benjamin Fanjoy/Getty Images).
      • Welcome back to This Week in Stratechery!
      • As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone.
      • Additionally, you have complete control over what we send to you.
      • On that note, here were a few of our favorites this week.
    • EN 要点:
      • (Photo by Benjamin Fanjoy/Getty Images)
      • Welcome back to This Week in Stratechery
      • As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone
      • Additionally, you have complete control over what we send to you

ArXiv cs.AI (B_intro+search) 链接到标题

  • Regularized Emphatic Temporal-Difference Learning: Stability under Constant Stepsizes

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19170v1 Announce Type: new.
      • Abstract: Emphatic temporal-difference learning (ETD) stabilizes the expected off-policy TD update and changes its projection geometry, but neither property determines constant-stepsize sampled dynamics.
      • We construct an ergodic two-state counterexample in which the ETD mean map contracts while the sampled product has a positive top Lyapunov exponent.
      • Regenerative-cycle analysis separates this sign from the infinite variance of the follow-on trace.
    • EN 要点:
      • arXiv:2609.19170v1 Announce Type: new
      • Abstract: Emphatic temporal-difference learning (ETD) stabilizes the expected off-policy TD update and changes its projection geometry, but neither property det…
      • We construct an ergodic two-state counterexample in which the ETD mean map contracts while the sampled product has a positive top Lyapunov exponent
      • Regenerative-cycle analysis separates this sign from the infinite variance of the follow-on trace

  • BioPhys-Bridge: A Benchmark for Interdisciplinary Scientific Reasoning in Physics-Grounded Biological Research

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19180v1 Announce Type: new.
      • Abstract: Language models face unique challenges in analyzing interdisciplinary scientific research literature.
      • In biophysics research, faithful answers require grounding observed data in source evidence, interpreting it through a quantitative physics model, and linking it to a biological mechanism.
      • To address this challenge, we introduce BioPhys-Bridge, a novel benchmark dataset for evidence-grounded scientific reasoning over biophysical literature.
    • EN 要点:
      • arXiv:2609.19180v1 Announce Type: new
      • Abstract: Language models face unique challenges in analyzing interdisciplinary scientific research literature
      • In biophysics research, faithful answers require grounding observed data in source evidence, interpreting it through a quantitative physics model, and linking i…
      • To address this challenge, we introduce BioPhys-Bridge, a novel benchmark dataset for evidence-grounded scientific reasoning over biophysical literature

  • What Do We Expect from LLMs? Mapping the Design of LLM Benchmarks

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19182v1 Announce Type: new.
      • Abstract: Benchmarks are central to how progress in large language models (LLMs) is assessed and communicated.
      • Yet model rankings alone reveal little about how evaluation requirements themselves are changing.
      • The expanding variety of benchmarks offers another perspective: what researchers expect LLMs to do, and what they count as successful performance.
    • EN 要点:
      • arXiv:2609.19182v1 Announce Type: new
      • Abstract: Benchmarks are central to how progress in large language models (LLMs) is assessed and communicated
      • Yet model rankings alone reveal little about how evaluation requirements themselves are changing
      • The expanding variety of benchmarks offers another perspective: what researchers expect LLMs to do, and what they count as successful performance

  • Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19203v1 Announce Type: new.
      • Abstract: AI applications have shifted from single, monolithic foundation models (FM) to compound agentic systems.
      • Yet today’s stacks remain fragmented: even as protocols (e.g., MCP, A2A) ease tool/agent connectivity, each framework embeds an implicit runtime for state, memory, budgets, and guardrails, making behavior non-portable and governance brittle.
      • It mirrors computing before operating systems, when every program re-implemented basic services.
    • EN 要点:
      • arXiv:2609.19203v1 Announce Type: new
      • Abstract: AI applications have shifted from single, monolithic foundation models (FM) to compound agentic systems
      • Yet today’s stacks remain fragmented: even as protocols (e.g., MCP, A2A) ease tool/agent connectivity, each framework embeds an implicit runtime for state, memo…
      • It mirrors computing before operating systems, when every program re-implemented basic services

  • What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19212v1 Announce Type: new.
      • Abstract: Systematic generalization, the ability to solve novel problems by recombining known atomic elements, is central to human intelligence but difficult to study rigorously under controlled settings.
      • Existing studies therefore rely on simplifications such as approximately linear action composition, productivity-based tests, and action-explicit goals, which make systematic generalization easier to study but omit some essential aspects of this capability.
      • To characterize what these simplifications miss, we adopt a reasoning-centered lens and introduce TranSGrid, a testbed that brings deductive, inductive, and abductive reasoning together within a unified task.
    • EN 要点:
      • arXiv:2609.19212v1 Announce Type: new
      • Abstract: Systematic generalization, the ability to solve novel problems by recombining known atomic elements, is central to human intelligence but difficult to…
      • Existing studies therefore rely on simplifications such as approximately linear action composition, productivity-based tests, and action-explicit goals, which m…
      • To characterize what these simplifications miss, we adopt a reasoning-centered lens and introduce TranSGrid, a testbed that brings deductive, inductive, and abd…

  • Characterizing Web Search by Conversational LLM Agents: From Search Decisions and Strategies to Results and Responses

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19244v1 Announce Type: new.
      • Abstract: Conversational LLM agents increasingly rely on Web search, yet the end-to-end lifecycle of agentic search remains poorly understood.
      • We present the first study of Web search across four major conversational platforms (ChatGPT, Claude, Grok, and DeepSeek), combining real-world user interactions (invivo) with controlled experiments using the same platform’s models by their APIs (invitro).
      • We investigate the quality of agentic decisions to invoke Web search, their strategies to formulate queries, the potential domain preferences in the search results they receive, and the choices they make when transforming search results into grounded responses.
    • EN 要点:
      • arXiv:2609.19244v1 Announce Type: new
      • Abstract: Conversational LLM agents increasingly rely on Web search, yet the end-to-end lifecycle of agentic search remains poorly understood
      • We present the first study of Web search across four major conversational platforms (ChatGPT, Claude, Grok, and DeepSeek), combining real-world user interaction…
      • We investigate the quality of agentic decisions to invoke Web search, their strategies to formulate queries, the potential domain preferences in the search resu…

  • Do AI Agents Understand Computer Architecture?

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19387v1 Announce Type: new.
      • Abstract: Agents are increasingly asked to design hardware, and increasingly reported to succeed.
      • Such reports establish that a design improved; they cannot establish why.
      • An agent that improves an accelerator may be reasoning about the machine, or may be searching competently over knobs whose meaning it never recovers – and only the first transfers to the next architecture.
    • EN 要点:
      • arXiv:2609.19387v1 Announce Type: new
      • Abstract: Agents are increasingly asked to design hardware, and increasingly reported to succeed
      • Such reports establish that a design improved; they cannot establish why
      • An agent that improves an accelerator may be reasoning about the machine, or may be searching competently over knobs whose meaning it never recovers – and only…

  • MAGS: Multi-agent Auto-formalization Guarantees Safety for Agentic Outputs

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19391v1 Announce Type: new.
      • Abstract: LLM coding agents now generate complex programs at a scale that makes thorough human review increasingly difficult, raising the risk of safety and security failures.
      • Common approaches, including fuzz testing, static analysis, and LLM-as-a-Verifier, can detect many failures but struggle to cover all possible edge cases.
      • Formal verification addresses this by providing machine-checkable guarantees over specified properties, but traditionally demands substantial manual specification and proof engineering.
    • EN 要点:
      • arXiv:2609.19391v1 Announce Type: new
      • Abstract: LLM coding agents now generate complex programs at a scale that makes thorough human review increasingly difficult, raising the risk of safety and sec…
      • Common approaches, including fuzz testing, static analysis, and LLM-as-a-Verifier, can detect many failures but struggle to cover all possible edge cases
      • Formal verification addresses this by providing machine-checkable guarantees over specified properties, but traditionally demands substantial manual specificati…

  • Closed-World Resolution Against Tool Hallucination in LLM Agents

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19425v1 Announce Type: new.
      • Abstract: Tool-augmented large language model (LLM) agents fail in a way no tool-selection or tool-security method addresses: they call tools that do not exist and pass arguments no schema declares.
      • Existing defenses either pick the right tool (selection) or constrain what an agent may do with real tools (gating), both of which presuppose the emitted call refers to a real tool at all.
      • We show this is a structural blind spot: a hallucinated call is by construction not a decision any gate made, so no gate can reject it.
    • EN 要点:
      • arXiv:2609.19425v1 Announce Type: new
      • Abstract: Tool-augmented large language model (LLM) agents fail in a way no tool-selection or tool-security method addresses: they call tools that do not exist…
      • Existing defenses either pick the right tool (selection) or constrain what an agent may do with real tools (gating), both of which presuppose the emitted call r…
      • We show this is a structural blind spot: a hallucinated call is by construction not a decision any gate made, so no gate can reject it

  • The syntax and semantics of goals

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19448v1 Announce Type: new.
      • Abstract: In both cognitive science and computer science, goals are conceptualized as cognitive states that flexibly combine with world knowledge to organize and specify purposeful behavior.
      • In this way, goals are compositional representations whose content relates to rational behavior.
      • We here draw attention to goals as representations and their content because it highlights a parallel with other areas in cognitive science - in particular, the syntax-semantics interface in linguistics and logic - while also foregrounding foundational questions about the expressivity, design, and efficiency of different goal representations.
    • EN 要点:
      • arXiv:2609.19448v1 Announce Type: new
      • Abstract: In both cognitive science and computer science, goals are conceptualized as cognitive states that flexibly combine with world knowledge to organize an…
      • In this way, goals are compositional representations whose content relates to rational behavior
      • We here draw attention to goals as representations and their content because it highlights a parallel with other areas in cognitive science - in particular, the…

ArXiv cs.CL (B_intro+search) 链接到标题

  • Modality Discrepancy Transformer for Ambivalence and Hesitancy Recognition

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19148v1 Announce Type: new.
      • Abstract: Ambivalence and hesitancy (A/H) are affective states in which individuals express contradictory signals across facial, vocal, and linguistic channels.
      • Automatically recognising A/H in clinical videos requires detecting cross-modal disagreement – the signal that standard fusion methods suppress.
      • Based on the conflict-aware multimodal fusion framework of Bekhouche et al., we present the Modality Discrepancy Transformer (MDT).
    • EN 要点:
      • arXiv:2609.19148v1 Announce Type: new
      • Abstract: Ambivalence and hesitancy (A/H) are affective states in which individuals express contradictory signals across facial, vocal, and linguistic channels
      • Automatically recognising A/H in clinical videos requires detecting cross-modal disagreement – the signal that standard fusion methods suppress
      • Based on the conflict-aware multimodal fusion framework of Bekhouche et al., we present the Modality Discrepancy Transformer (MDT)

  • Subliminal Prompting Beyond Static Geometry: Causal Depth and Multi-Token Confounds

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19149v1 Announce Type: new.
      • Abstract: Subliminal learning shows that language models can transmit a hidden trait through outputs that appear unrelated to it.
      • One proposed explanation, token entanglement, links animal and number tokens through the model’s output vocabulary.
      • Yet existing measurements answer different questions: whether outputs co-vary, fixed output vectors align, an answer can be read from a hidden state, or that state causally controls the answer.
    • EN 要点:
      • arXiv:2609.19149v1 Announce Type: new
      • Abstract: Subliminal learning shows that language models can transmit a hidden trait through outputs that appear unrelated to it
      • One proposed explanation, token entanglement, links animal and number tokens through the model’s output vocabulary
      • Yet existing measurements answer different questions: whether outputs co-vary, fixed output vectors align, an answer can be read from a hidden state, or that st…

  • Sampling Reveals Style: Unsupervised, Training-Free Discovery of Prompt-Conditional Stylistic Axes in LLM Activations

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19150v1 Announce Type: new.
      • Abstract: Large language models (LLMs) encode rich stylistic structure in their hidden activations, but discovering which stylistic dimensions are salient for a given prompt typically requires supervised contrastive data.
      • We present a training-free, prompt-conditional alternative: we repeatedly sample completions of a single prompt at elevated temperature, apply Principal Component Analysis (PCA) to the pooled hidden activations, and label the resulting axes automatically from the pole generations.
      • We validate the discovered axes against 245 human-elicited stylistic annotations in a two-phase study.
    • EN 要点:
      • arXiv:2609.19150v1 Announce Type: new
      • Abstract: Large language models (LLMs) encode rich stylistic structure in their hidden activations, but discovering which stylistic dimensions are salient for a…
      • We present a training-free, prompt-conditional alternative: we repeatedly sample completions of a single prompt at elevated temperature, apply Principal Compone…
      • We validate the discovered axes against 245 human-elicited stylistic annotations in a two-phase study

  • What Users Think of Generative AI: A Cross-Platform NLP Analysis of Trust and Friction in App Store Reviews

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19151v1 Announce Type: new.
      • Abstract: Generative AI (GenAI) applications have achieved rapid consumer adoption, yet little large-scale research examines user-perceived quality, trust, and adoption barriers.
      • We present one of the first cross-application analyses of app store reviews for six major GenAI applications (ChatGPT, Gemini, Microsoft Copilot, Claude, DeepSeek, and Perplexity), comprising 17,012 English-language reviews from Google Play and the Apple App Store.
      • We combine BERTopic topic modeling with RoBERTa sentiment classification and evaluate cross-application differences using chi-square, Kruskal-Wallis, and multinomial logistic regression with Bonferroni correction.
    • EN 要点:
      • arXiv:2609.19151v1 Announce Type: new
      • Abstract: Generative AI (GenAI) applications have achieved rapid consumer adoption, yet little large-scale research examines user-perceived quality, trust, and…
      • We present one of the first cross-application analyses of app store reviews for six major GenAI applications (ChatGPT, Gemini, Microsoft Copilot, Claude, DeepSe…
      • We combine BERTopic topic modeling with RoBERTa sentiment classification and evaluate cross-application differences using chi-square, Kruskal-Wallis, and multin…

  • FakeSpotter: A content and strategy agnostic Viral Misinformation Detection Tool

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19152v1 Announce Type: new.
      • Abstract: Misinformation detection tools often rely on binary true and false classifications or models trained on historical examples, limiting their usefulness when novel misleading narratives emerge.
      • Here, we present FakeSpotter, a content- and strategy-agnostic tool designed to estimate the viral misinformation risk of textual content by measuring structural fingerprints of misinformation rather than directly adjudicating truthfulness.
      • FakeSpotter operationalizes a theory-driven framework across linguistic, narrative, logical, and critical-thinking dimensions, using repeated LLM assessments and domain-specific logistic regression classifiers for short and long texts.
    • EN 要点:
      • arXiv:2609.19152v1 Announce Type: new
      • Abstract: Misinformation detection tools often rely on binary true and false classifications or models trained on historical examples, limiting their usefulness…
      • Here, we present FakeSpotter, a content- and strategy-agnostic tool designed to estimate the viral misinformation risk of textual content by measuring structura…
      • FakeSpotter operationalizes a theory-driven framework across linguistic, narrative, logical, and critical-thinking dimensions, using repeated LLM assessments an…

  • Stop Removing Stopwords: How an Inherited Preprocessing Default Distorts Legal Text-as-Data

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19153v1 Announce Type: new.
      • Abstract: Empirical legal scholarship increasingly treats judicial text as data, and much of it still runs on sparse, interpretable pipelines – TF-IDF features and linear classifiers – because the textual feature is often the object of study, not merely a means to a prediction.
      • Yet these pipelines inherit a chain of preprocessing defaults from mid-century information retrieval that were never validated against classification accuracy, the most entrenched being stopword removal.
      • This study introduces an exhaustive single-word ablation that measures a preprocessing step’s effect directly against the downstream objective, and applies it to stopword removal as the hardest case to dislodge.
    • EN 要点:
      • arXiv:2609.19153v1 Announce Type: new
      • Abstract: Empirical legal scholarship increasingly treats judicial text as data, and much of it still runs on sparse, interpretable pipelines – TF-IDF features…
      • Yet these pipelines inherit a chain of preprocessing defaults from mid-century information retrieval that were never validated against classification accuracy,…
      • This study introduces an exhaustive single-word ablation that measures a preprocessing step’s effect directly against the downstream objective, and applies it t…

  • Neo-Classic: A Benchmark for Evaluating Linguistic-Aesthetic Reasoning in Classical Chinese Poetry

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19154v1 Announce Type: new.
      • Abstract: While Large Language Models (LLMs) achieve high accuracy on established Classical Chinese Poetry benchmarks, it remains challenging to distinguish transferable Linguistic-Aesthetic Reasoning from reliance on familiar pre-training patterns.
      • To address this issue, we introduce Neo-Classic, an evaluation benchmark that combines a constructionist Out-of-Sample (OOS) dataset with a suite of reverse understanding probes.
      • Unlike traditional benchmarks that rely on verification or generation over historical corpora, Neo-Classic comprises strictly metrical poetry authored by contemporary experts, reducing the possibility of direct retrieval.
    • EN 要点:
      • arXiv:2609.19154v1 Announce Type: new
      • Abstract: While Large Language Models (LLMs) achieve high accuracy on established Classical Chinese Poetry benchmarks, it remains challenging to distinguish tra…
      • To address this issue, we introduce Neo-Classic, an evaluation benchmark that combines a constructionist Out-of-Sample (OOS) dataset with a suite of reverse und…
      • Unlike traditional benchmarks that rely on verification or generation over historical corpora, Neo-Classic comprises strictly metrical poetry authored by contem…

  • Towards Proactive Detection of User-Side Implicit Conflicts in Human-LLM Dialogue

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19155v1 Announce Type: new.
      • Abstract: In Human-LLM dialogue, follow-up user utterances may implicitly conflict with earlier intents, leading the LLM to misinterpret user needs and generate inappropriate responses.
      • A reliable dialogue system should proactively detect user-side conflicts before generating a response and seek clarification when necessary.
      • However, prior work has largely focused on LLM-side conflicts, leaving user-side conflicts underexplored.
    • EN 要点:
      • arXiv:2609.19155v1 Announce Type: new
      • Abstract: In Human-LLM dialogue, follow-up user utterances may implicitly conflict with earlier intents, leading the LLM to misinterpret user needs and generate…
      • A reliable dialogue system should proactively detect user-side conflicts before generating a response and seek clarification when necessary
      • However, prior work has largely focused on LLM-side conflicts, leaving user-side conflicts underexplored

  • Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from Mistakes

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19156v1 Announce Type: new.
      • Abstract: Data-driven fine-tuning is widely adopted to enhance reasoning in Large Language Models (LLMs) due to its simplicity and efficiency.
      • However, mainstream imitation learning methods that rely exclusively on perfect reasoning trajectories suffer from a Scaling Collapse: when the problem set is limited, increasing positive examples fails to yield continuous improvement.
      • However, during inference, an LLM can not guarantee that every intermediate step is correct and is therefore prone to errors.
    • EN 要点:
      • arXiv:2609.19156v1 Announce Type: new
      • Abstract: Data-driven fine-tuning is widely adopted to enhance reasoning in Large Language Models (LLMs) due to its simplicity and efficiency
      • However, mainstream imitation learning methods that rely exclusively on perfect reasoning trajectories suffer from a Scaling Collapse: when the problem set is l…
      • However, during inference, an LLM can not guarantee that every intermediate step is correct and is therefore prone to errors

  • VisKG-LM: Compiling Knowledge Graphs into Visual Memory for Multiple-Choice Question Answering

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19158v1 Announce Type: new.
      • Abstract: Knowledge graphs are usually integrated into question answering by encoding a retrieved subgraph with a graph neural network and fusing it with the language model in the online inference path.
      • The same subgraph is therefore re-encoded from scratch every time a pair is scored, across training epochs, seeds, and evaluation runs, even though the knowledge graph never changes.
      • We ask whether the retrieved knowledge graphs can instead be compiled once, offline, and then accessed as read-only memory.
    • EN 要点:
      • arXiv:2609.19158v1 Announce Type: new
      • Abstract: Knowledge graphs are usually integrated into question answering by encoding a retrieved subgraph with a graph neural network and fusing it with the la…
      • The same subgraph is therefore re-encoded from scratch every time a pair is scored, across training epochs, seeds, and evaluation runs, even though the knowledg…
      • We ask whether the retrieved knowledge graphs can instead be compiled once, offline, and then accessed as read-only memory

ArXiv cs.LG (B_intro+search) 链接到标题

  • Generative Query Suggestion via Intent Coverage and Query-Level Credit Assignment

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19209v1 Announce Type: new.
      • Abstract: Generative query suggestion aims to enhance user engagement by anticipating user intents and recommending relevant follow-up queries.
      • A central challenge is to generate slates whose individual queries are useful while the slate covers distinct intents.
      • We propose an Intent-Driven Query Suggestion Framework with dual-stage optimization.
    • EN 要点:
      • arXiv:2609.19209v1 Announce Type: new
      • Abstract: Generative query suggestion aims to enhance user engagement by anticipating user intents and recommending relevant follow-up queries
      • A central challenge is to generate slates whose individual queries are useful while the slate covers distinct intents
      • We propose an Intent-Driven Query Suggestion Framework with dual-stage optimization

  • Layer-wise Curriculum Learning for Efficient LLM Compression

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19213v1 Announce Type: new.
      • Abstract: In this paper, we introduce layer-wise curriculum learning for efficient LLM compression.
      • The proposed method facilitates the knowledge transfer from the teacher model to the student model, utilizing a curriculum learning approach that begins with easier optimization tasks and progressively tackles harder ones.
      • In order to adopt the layer-wise learning in LLM compression, we partition the whole model into multiple segments consisting of layers, thereby enabling more computationally efficient knowledge transfer for LLMs.
    • EN 要点:
      • arXiv:2609.19213v1 Announce Type: new
      • Abstract: In this paper, we introduce layer-wise curriculum learning for efficient LLM compression
      • The proposed method facilitates the knowledge transfer from the teacher model to the student model, utilizing a curriculum learning approach that begins with ea…
      • In order to adopt the layer-wise learning in LLM compression, we partition the whole model into multiple segments consisting of layers, thereby enabling more co…

  • Block Parallelism For Efficient Distributed Long-Context Diffusion Language Model Training

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19242v1 Announce Type: new.
      • Abstract: Block diffusion language models (BDLMs) combine autoregressive dependencies across blocks with parallel denoising within blocks, but long-context training is constrained by distributed attention communication and activation memory.
      • Conventional context parallelism (CP) shards the combined clean-plus-corrupted sequence by position, communicating shared clean K/V together with block-specific corrupted K/V and their gradients.
      • We observe that the BDLM objective separates over target blocks.
    • EN 要点:
      • arXiv:2609.19242v1 Announce Type: new
      • Abstract: Block diffusion language models (BDLMs) combine autoregressive dependencies across blocks with parallel denoising within blocks, but long-context trai…
      • Conventional context parallelism (CP) shards the combined clean-plus-corrupted sequence by position, communicating shared clean K/V together with block-specific…
      • We observe that the BDLM objective separates over target blocks

  • Randomized SVD Approximations for Spectral Co-Clustering of Word-Document Matrices

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19243v1 Announce Type: new.
      • Abstract: Spectral co-clustering is a useful tool for discovering latent structure in word-document matrices, but its reliance on singular value decomposition (SVD) can make standard formulations expensive on high-dimensional data.
      • This paper presents two randomized approximations for normalized spectral co-clustering of bipartite text data when the numbers of document and word clusters may differ.
      • The first method uses randomized SVD through random projection, while the second combines partial SVD with element-wise random sampling.
    • EN 要点:
      • arXiv:2609.19243v1 Announce Type: new
      • Abstract: Spectral co-clustering is a useful tool for discovering latent structure in word-document matrices, but its reliance on singular value decomposition (…
      • This paper presents two randomized approximations for normalized spectral co-clustering of bipartite text data when the numbers of document and word clusters ma…
      • The first method uses randomized SVD through random projection, while the second combines partial SVD with element-wise random sampling

  • Radio-Frequency Convolutional Neural Networks

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19279v1 Announce Type: new.
      • Abstract: Running artificial intelligence (AI) models directly on edge devices such as smartphones, wearables, and drones offers low latency, pervasive scalability, and data privacy, but these devices rarely carry the computing capability that modern neural networks demand.
      • Edge accelerators have been developed in response, yet each adds computing hardware to devices already constrained in size, weight, power, and cost (SWaP-C).
      • An alternative lies in what these devices already carry: the frequency mixer in every wireless radio multiplies signals in time, natively performing convolution in the frequency domain.
    • EN 要点:
      • arXiv:2609.19279v1 Announce Type: new
      • Abstract: Running artificial intelligence (AI) models directly on edge devices such as smartphones, wearables, and drones offers low latency, pervasive scalabil…
      • Edge accelerators have been developed in response, yet each adds computing hardware to devices already constrained in size, weight, power, and cost (SWaP-C)
      • An alternative lies in what these devices already carry: the frequency mixer in every wireless radio multiplies signals in time, natively performing convolution…

  • Learning-Induced Dynamical Transition in Recurrent Neural Networks

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19288v1 Announce Type: new.
      • Abstract: Learning in recurrent neural networks can fundamentally reshape their underlying dynamics, transforming initially chaotic activity into stable task-dependent behavior.
      • We develop a non-equilibrium dynamical mean-field theory(DMFT) to describe this transition during learning.
      • We show that a slow feedback-driven learning process generates an evolving effective feedback strength that drives the network through a transition from chaotic to stable dynamics defined by a bifurcation of the DMFT solution.
    • EN 要点:
      • arXiv:2609.19288v1 Announce Type: new
      • Abstract: Learning in recurrent neural networks can fundamentally reshape their underlying dynamics, transforming initially chaotic activity into stable task-de…
      • We develop a non-equilibrium dynamical mean-field theory(DMFT) to describe this transition during learning
      • We show that a slow feedback-driven learning process generates an evolving effective feedback strength that drives the network through a transition from chaotic…

  • Personalized Federated Hierarchical Gaussian Processes for Privacy-Preserving Modeling of Heterogeneous Distributed Systems

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19337v1 Announce Type: new.
      • Abstract: We present Personalized Federated Hierarchical Gaussian Processes (pFedHGP) for probabilistic regression and classification when data are distributed across heterogeneous clients.
      • Each client’s latent function decomposes into (i) a shared global component, (ii) a client-specific deviation that shares the global kernel structure, and (iii) a flexible local residual.
      • Sparse inducing-variable approximations and federated variational inference keep raw data local while the server synchronizes only low-dimensional statistics for the shared component.
    • EN 要点:
      • arXiv:2609.19337v1 Announce Type: new
      • Abstract: We present Personalized Federated Hierarchical Gaussian Processes (pFedHGP) for probabilistic regression and classification when data are distributed…
      • Each client’s latent function decomposes into (i) a shared global component, (ii) a client-specific deviation that shares the global kernel structure, and (iii)…
      • Sparse inducing-variable approximations and federated variational inference keep raw data local while the server synchronizes only low-dimensional statistics fo…

  • How to Guide Your Language Flow

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19356v1 Announce Type: new.
      • Abstract: We introduce a new method to guide flow matching models.
      • Our approach, which we call probe guidance, uses the frozen internal states of an existing diffusion model to construct a guidance signal.
      • This works using a similar principle as autoguidance, but eliminates the need for an additional forward pass at inference time and provides a reliable path to ensure that the weak and strong model share similar dynamics.
    • EN 要点:
      • arXiv:2609.19356v1 Announce Type: new
      • Abstract: We introduce a new method to guide flow matching models
      • Our approach, which we call probe guidance, uses the frozen internal states of an existing diffusion model to construct a guidance signal
      • This works using a similar principle as autoguidance, but eliminates the need for an additional forward pass at inference time and provides a reliable path to e…

  • Smart Insole Human Activity Recognition for Continuous Monitoring in Elderly Care

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19359v1 Announce Type: new.
      • Abstract: Falls in older adults are often preceded by changes in mobility, balance, and postural transitions.
      • This paper presents a wireless smart insole platform and machine-learning workflow for recognizing sitting, standing, walking, and unstable walking from plantar-pressure and inertial signals.
      • Each insole integrates 16 active pressure-sensing locations and a six-dimensional IMU stream consisting of tri-axial acceleration and angular velocity.
    • EN 要点:
      • arXiv:2609.19359v1 Announce Type: new
      • Abstract: Falls in older adults are often preceded by changes in mobility, balance, and postural transitions
      • This paper presents a wireless smart insole platform and machine-learning workflow for recognizing sitting, standing, walking, and unstable walking from plantar…
      • Each insole integrates 16 active pressure-sensing locations and a six-dimensional IMU stream consisting of tri-axial acceleration and angular velocity

  • Stiefel Attention: When the Geometry of Transformer Projection Matrices Dominates Optimizer Choice—and When It Does Not

    • 发布时间:2026-09-18 12:00 北京时间
    • 摘要:【待翻译】- arXiv:2609.19363v1 Announce Type: new.
      • Abstract: The query and key projections $\WQ,\WK$ in attention are almost always trained by Euclidean optimizers with no constraint on their geometry.
      • We constrain them to the Stiefel manifold and optimize them there with a Riemannian Adam that carries one scalar second moment per frame, caps its step by a trust region, and retracts polarly.
      • Four propositions prove this update is steepest descent in the embedded metric, independent of gradient scale, well conditioned, and exactly $\mathrm{O}(d)$-equivariant, each certified numerically in \texttt{float64}.
    • EN 要点:
      • arXiv:2609.19363v1 Announce Type: new
      • Abstract: The query and key projections $\WQ,\WK$ in attention are almost always trained by Euclidean optimizers with no constraint on their geometry
      • We constrain them to the Stiefel manifold and optimize them there with a Riemannian Adam that carries one scalar second moment per frame, caps its step by a tru…
      • Four propositions prove this update is steepest descent in the embedded metric, independent of gradient scale, well conditioned, and exactly $\mathrm{O}(d)$-equ…