System translated (Gemini)

🤖 AI 速览

OpenAI focuses on releasing GPT-6 Astra enterprise use cases: Harvey optimizes legal documents, invideo triples color grading efficiency, and Ringg customer service resolution rate reaches 65%. Security governance simultaneously heats up: Altman emphasizes human control at the UN Security Council, …
📋 文章元数据
发布时间
2026-09-24
类型
ai-daily
字数
7112
阅读时长
34 min

2026-09-24 AI Daily | AI Governance and Vertical Adoption Heat Up: Altman Discusses Human Control at UN Security Council, Astra Enters Legal, Color Grading, and Customer Service Link to heading

OpenAI releases a slate of GPT-6 Astra enterprise use cases: Harvey optimizes legal documents, invideo triples color grading efficiency, and Ringg achieves a 65% customer service resolution rate. Safety and governance efforts are also ramping up: Altman emphasizes human control at the UN Security Council, Google introduces server-side memory for Private AI Compute, and OpenAI expands cyber defense for Ukraine. On the evaluation front, audits are beginning for shortcut learning and numerical invariance.

📖 In-Depth Guide to This Issue’s Watch List Link to heading

Three main themes are worth watching today. First, enterprise-level AI adoption: OpenAI has released a set of GPT-6 Astra use cases, with Harvey improving the structuring of legal documents, invideo tripling color grading efficiency, and Ringg reaching up to a 65% customer service resolution rate. Gemini 3.8 TTS is also worth a close look for voice and engineering teams. Second, safety and governance: Google is introducing server-side memory for Private AI Compute, Altman discussed human control and international collaboration at the UN Security Council, and OpenAI has also released MentalHealthBench and expanded its civil cyber defense efforts in Ukraine—recommended for policy and security teams to follow. Third, reflections on evaluation and training: two arXiv papers audit shortcut learning in fake news datasets and test numerical representation invariance, respectively. There is also a paper on end-to-end training experience in Rust, which is recommended for in-depth reading by model evaluation and infrastructure teams.

🌐 AI Hot Topics on X Link to heading

Topic 1: Meta Launches Muse AI Agent Amid Open-Source Inspiration and Logo Memes Link to heading

  • Category: AI · News
  • Summary: Trending for: 1 day ago, Related posts: 4900
  • What happened: Meta released an AI agent named Muse, accompanied by discussions on open-source inspiration and logo memes.
  • Why it’s important: This move shows Meta is doubling down on the AI agent race, and its open-source strategy could impact the industry’s ecosystem and competitive landscape.
  • Discussion summary: Discussions on X are mainly focused on whether it’s truly open-source, how its capabilities compare to other AI agents, and the memes and brand promotion surrounding its logo.

Topic 2: Meta’s Alexandr Wang Embraces Butthole Logo Memes for New AI Agent Muse Link to heading

  • Category: AI · Other
  • Summary: Trending for: 1 day ago, Related posts: 4800
  • What happened: After the release of Meta’s new AI agent, Muse, its logo was mocked by netizens as a “butthole” design. Alexandr Wang publicly embraced and responded to the meme.
  • Why it’s important: This incident shows that AI product marketing is becoming deeply intertwined with meme culture. A controversial visual symbol helped the new agent gain high visibility on social media, but it also tests the brand’s ability to steer public attention from mockery to the product’s actual capabilities.
  • Discussion summary: The main debates on X revolve around whether the logo is inappropriate, whether embracing the meme is clever or awkward, and whether the meme topic has diluted the focus on Muse’s actual performance.

Topic 3: Meta’s Muse Mascot Steals Spotlight Ahead of Connect Link to heading

  • Category: AI · News
  • Summary: Trending for: 7 hours ago, Related posts: 762
  • What happened: Meta’s Muse mascot became a focal point before the Connect conference due to an AI-generated video.
  • Why it’s important: This showcases the creative potential of generative AI video tools in brand and IP promotion. It also serves as a warm-up for Meta Connect and highlights the growing attention on multi-modal AI creation platforms.
  • Discussion summary: Discussions on X primarily cover the generation quality and choice of different AI video models, potential announcements at Meta Connect, and whether all-in-one creation platforms that aggregate multiple models have a competitive edge.

Topic 4: Alexandr Wang Hypes Meta’s Muse AI with Playful Mascot Posts Link to heading

  • Category: AI · Other
  • Summary: Trending for: 17 hours ago, Related posts: 1900
  • What happened: Scale AI founder Alexandr Wang promoted Meta’s Muse AI on X by posting playful mascot-related content.
  • Why it’s important: This action shows a key figure in the AI industry publicly endorsing Meta’s new model, which could boost Muse AI’s visibility and credibility within the developer community and intensify competition among generative AI tools.
  • Discussion summary: The focus of discussion includes Muse AI’s actual capabilities and product positioning, whether Alexandr Wang is just memeing or if there’s a business collaboration, and whether the model can differentiate itself from existing image or video generation tools.

Topic 5: Katzenberg Urges Hollywood to Embrace AI as Creative Tool Link to heading

  • Category: AI · News
  • Overview: Trending Time: 7 hours ago, Related Posts: 5200
  • What it is: Jeffrey Katzenberg has publicly urged Hollywood to embrace artificial intelligence as a creative tool, rather than viewing it as a threat.
  • Why it matters: The push by Hollywood executives to integrate generative AI into the core creative process could accelerate its mainstream adoption in film and television screenwriting, special effects, and content production.
  • Discussion overview: The discussion on X focuses on whether AI will replace creative jobs, how copyright and compensation will be defined, and whether AI can truly be a creative partner rather than undermining originality.

Topic 6: AI Video Shows Trump as Mexican Grandmother in Kitchen Parody Link to heading

  • Category: AI · Other
  • Overview: Trending Time: , Related Posts: 707
  • What it is: An AI-generated video portrays Trump as a Mexican grandmother in a kitchen parody.
  • Why it matters: The video highlights the capabilities and potential risks of AI-generated video in imitating political figures, transforming cultural identities, and creating deepfakes.
  • Discussion overview: Discussions on X are centered on whether this is harmless entertainment parody or a misleading deepfake, and whether AI-generated content featuring political figures requires stricter labeling and regulation.

Today’s AI Public Opinion Summary on X Link to heading

Today’s main narrative revolves around Meta’s release of the Muse AI agent and the parody of its logo. The consensus is that AI product marketing is now deeply intertwined with meme culture. Controversial symbols and endorsements from key figures can quickly generate high exposure and intensify competition in the generative AI space. Points of contention primarily focus on whether Muse is truly open-source, whether the logo controversy and embrace of memes are clever or awkward, whether Alexandr Wang’s involvement is a joke or a business collaboration, and whether Hollywood should view AI as a creative tool or a threat to creative jobs. Potential risks include AI-generated parodies of political figures potentially crossing the line into misleading deepfakes, the lack of clear definitions for copyright and creator compensation, and the risk of brand attention being diluted by meme topics, causing the product’s actual capabilities to be overlooked.

💡 Influencer Insights Link to heading

No influencer insights for today. We recommend reading the in-depth content from the Watch List.

📚 Appendix: Today’s Watch List Update Source List Link to heading

Timeframe: Last 3 days; 22 sources covered; 43 updates in total

All-In Podcast (A_full) Link to heading

  • Steve Hilton & Spencer Pratt: Fixing California, Cheaper Gas, Ballot Fraud & CA Republicans
    • Published: 2026-09-23 09:44 Beijing Time
    • Abstract: - IREN is a vertically integrated AI cloud platform providing data centers, computing, and software for AI training and inference.
      • Oracle connects data, applications, and infrastructure to turn AI into business results, offering optimized flexibility, choice, and control as AI evolves.
      • Ernst & Young helps tech innovators from startups to exits to large corporations.
      • We are committed to providing everyone with the tools to reach their full potential and ensuring that all can benefit from technology.
      • Keel Infrastructure has the power, land, and connectivity required for high-performance computing and AI operations, backed by secure energy assets and grid interconnections across North America.
    • EN Key Points:
      • (0:00) Steve Hilton joins the Besties
      • (3:53) How one-party rule broke the state, regulatory bloat & who really writes the laws
      • (12:09) The path to victory: the persuadable voter, the cost-of-living pitch & the case against Becerra
      • (21:46) Spencer Pratt joins: the message that cut through & the LA race that flipped

Stratechery by Ben Thompson (A_full) Link to heading

  • More on Muse, Amazon, and Walmart; Muse and Expedia; Whither Google?
    • Published: 2026-09-23 18:00 Beijing Time
  • Summary:【待翻译】- Meta needs Walmart to wait out Amazon; Expedia seeks to keep its middleware position; meanwhile, where is Google?
    • $15 / month or $150 / year.
    • Substantial analysis of the news of the day delivered via three weekly emails or podcasts.
    • Stratechery Interviews.
    • Interviews with leading public CEOs, private company founders, and discussions with fellow analysts.
    • EN Highlights:
      • Meta needs Walmart to wait out Amazon; Expedia seeks to keep its middleware position; meanwhile, where is Google

OpenAI Blog (A_full) Link to heading

  • Two years of OpenAI Academy

    • Published: 2026-09-24 00:00 Beijing Time
    • Summary: - Marking two years of OpenAI Academy and bringing AI skills to even more communities.
      • This piece from OpenAI Blog explains how Two years of OpenAI Academy shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Two years of OpenAI Academy.
    • EN Highlights:
      • Marking two years of OpenAI Academy and bringing AI skills to even more communities.

  • OpenAI extends cyber access to Ukraine for civilian defense

    • Published: 2026-09-23 21:00 Beijing Time
    • Summary: - OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.
      • This piece from the OpenAI Blog explains how OpenAI’s extension of cyber access to Ukraine for civilian defense shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors after OpenAI extended cyber access to Ukraine for civilian defense.
    • EN Highlights:
      • OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.

  • Sam Altman’s remarks at the United Nations Security Council

    • Published: 2026-09-23 20:00 Beijing Time
    • Summary: - OpenAI CEO Sam Altman discusses AI safety, human control, and international cooperation in remarks to the United Nations Security Council.
      • This piece from OpenAI Blog explains how Sam Altman’s remarks at the United Nations Security Council shapes the broader AI and infrastructure landscape.

  • It also surfaces practical implications for founders, operators, and investors following Sam Altman’s remarks at the United Nations Security Council.

    • EN Summary:
      • OpenAI CEO Sam Altman discusses AI safety, human control, and international cooperation in remarks to the United Nations Security Council.

  • Harvey turns legal context into stronger drafts with GPT-6 Astra

    • Release Date: 2026-09-23 20:00 Beijing Time
    • Summary: GPT-6 Astra produces more structured, context-aware legal documents, freeing lawyers to focus on strategy.
      • This piece from OpenAI Blog explains how Harvey turns legal context into stronger drafts with GPT-6 Astra shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Harvey turns legal context into stronger drafts with GPT-6 Astra.
    • EN Summary:
      • GPT-6 Astra produces more structured, context-aware legal documents, freeing lawyers to focus on strategy.

  • How invideo improves color grading 3x with GPT‑6 Astra

    • Release Date: 2026-09-23 20:00 Beijing Time
    • Summary: With GPT‑6 Astra, invideo plans edits with greater precision, improves color correction and grading threefold, and produces 50 custom effects in one day.
      • This piece from OpenAI Blog explains how How invideo improves color grading 3x with GPT‑6 Astra shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following How invideo improves color grading 3x with GPT‑6 Astra.
    • EN Summary:
      • With GPT‑6 Astra, invideo plans edits with greater precision, improves color correction and grading threefold, and produces 50 custom effects in one day.

  • Ringg’s AI agents resolve up to 65% of customer calls with OpenAI

    • Release Date: 2026-09-23 20:00 Beijing Time

  • Summary: [To be translated] - Using GPT-5.6, Ringg powers multilingual agents across voice, chat, WhatsApp, and web for 90% less cost vs.

    • Using GPT-5.6, Ringg powers multilingual agents across voice, chat, WhatsApp, and web for 90% less cost vs.
    • Ringg’s AI agents resolve up to 65% of customer calls with OpenAI.
    • EN Key points:
      • Using GPT-5.6, Ringg powers multilingual agents across voice, chat, WhatsApp, and web for 90% less cost vs

  • Introducing MentalHealthBench

    • Posted: 2026-09-23 18:00 Beijing Time
    • Summary: - MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses in realistic mental health conversations.
      • This article from the OpenAI blog explains how the introduction of MentalHealthBench is shaping the broader AI and infrastructure landscape.
      • The introduction of MentalHealthBench also has practical implications for founders, operators, and investors.
    • EN Key points:
      • MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.

  • Airbnb widens access to GPT-6 Astra and OpenAI frontier models

    • Posted: 2026-09-23 09:00 Beijing Time
    • Summary: - Learn how Airbnb is expanding access to GPT-6 Astra and OpenAI’s frontier models to help engineering teams resolve bugs, design systems, and ship faster.
      • This article from the OpenAI blog explains how Airbnb’s expanded access to GPT-6 Astra and OpenAI’s frontier models are shaping the broader AI and infrastructure landscape.
      • Airbnb’s expansion of access to GPT-6 Astra and OpenAI’s frontier models also has practical implications for founders, operators, and investors.
    • EN Key points:
      • Learn how Airbnb is expanding access to GPT-6 Astra and OpenAI frontier models to help engineering teams solve bugs, design systems, and ship faster.

  • Grab and OpenAI bring practical AI skills to Southeast Asia

    • Posted: 2026-09-23 08:00 Beijing Time
    • Summary: - OpenAI and Grab are launching GO Forward with AI, a regional program to help 30,000 partners develop practical artificial intelligence skills across Southeast Asia.
      • This article from the OpenAI blog explains how Grab and OpenAI are bringing practical AI skills to Southeast Asia, shaping the broader AI and infrastructure landscape.
      • Grab and OpenAI bringing practical AI skills to Southeast Asia also has practical implications for founders, operators, and investors.
    • EN Key points:
      • OpenAI and Grab launch GO Forward with AI, a regional programme helping 30,000 partners build practical AI skills across Southeast Asia.

Google DeepMind Blog (A_full) Link to heading

  • Advancing Private AI Compute with secure, server-side memory

    • Publication Time: 2026-09-24 00:00 Beijing Time
    • Summary: [Translation Pending] - Introducing private, server-side memory to Private AI Compute for personal AI.
      • This piece from Google DeepMind Blog explains how Advancing Private AI Compute with secure, server-side memory shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Advancing Private AI Compute with secure, server-side memory.
    • EN Key Points:
      • Introducing private, server-side memory to Private AI Compute for personal AI.

  • Gemini 3.8 text-to-speech says hello

    • Publication Time: 2026-09-23 23:25 Beijing Time
    • Summary: [Translation Pending] - Gemini 3.8 text-to-speech says hello.
      • This piece from Google DeepMind Blog explains how Gemini 3.8 text-to-speech says hello shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Gemini 3.8 text-to-speech says hello.
    • EN Key Points:
      • Gemini 3.8 text-to-speech says hello

ArXiv cs.AI (B_intro+search) Link to heading

  • Do Synthetic Personas Predict Real Audience Response? A Sim-to-Real Study Where a No-Persona Baseline Beats Persona-Based Copy Simulation

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Summary: [Translation Pending] - arXiv:2609.25010v1 Announce Type: new.
      • Abstract: Marketers increasingly use large language models (LLMs) as “synthetic personas” to predict how an audience will react to a piece of copy before it ships, encouraged by evidence that profile-conditioned LLMs mimic human samples.
      • But is that prediction actually valid against real behaviour - and does the persona machinery help?
      • We present a sim-to-real validity study using the Upworthy Research Archive - thousands of headline A/B tests on shared real traffic, with measured click-through - as held-out ground truth.
    • EN Key Points:
      • arXiv:2609.25010v1 Announce Type: new

  • Abstract: Marketers increasingly use large language models (LLMs) as “synthetic personas” to predict how an audience will react to a piece of copy before it shi…

    • But is that prediction actually valid against real behaviour - and does the persona machinery help
    • We present a sim-to-real validity study using the Upworthy Research Archive - thousands of headline A/B tests on shared real traffic, with measured click-throug…

  • Do Existing Preconditioners Improve Biomedical Tabular Foundation Learning? An Empirical Study on TabPFN Optimization

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: [PENDING TRANSLATION] - arXiv:2609.25013v1 Announce Type: new.
      • Abstract: Tabular foundation models have recently shown strong potential for structured biomedical data analysis.
      • Among them, TabPFN has emerged as an effective approach for low-data tabular classification tasks.
      • However, the impact of optimization and preconditioning strategies on biomedical fine-tuning remains largely unexplored.
    • EN Highlights:
      • arXiv:2609.25013v1 Announce Type: new
      • Abstract: Tabular foundation models have recently shown strong potential for structured biomedical data analysis
      • Among them, TabPFN has emerged as an effective approach for low-data tabular classification tasks
      • However, the impact of optimization and preconditioning strategies on biomedical fine-tuning remains largely unexplored

  • 4DGS-JEPA: Temporally Compositional Joint-Embedding Prediction for Dynamic Gaussian Splatting

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25036v1 Announce Type: new.
      • Abstract: Dynamic Gaussian splatting provides an explicit representation of evolving 3D scenes, but existing methods are primarily optimized for reconstruction, future state generation, or rendering, rather than for learning reusable predictive dynamics.
      • We propose 4DGS-JEPA, a Gaussian-native joint-embedding predictive architecture for causal, multi-level prediction in dynamic Gaussian scenes.
      • The model uses hierarchical scene, motion group, and Gaussian-level representations along with a horizon-conditioned transition operator that supports both direct prediction and recursive rollout.
    • EN Highlights:
      • arXiv:2609.25036v1 Announce Type: new

  • Abstract: Dynamic Gaussian Splatting provides an explicit representation of evolving 3D scenes, but existing approaches are primarily optimized for reconstructi…

    • We propose 4DGS-JEPA, a Gaussian-native joint-embedding predictive architecture for causal multi-horizon prediction over dynamic Gaussian scenes
    • The model uses a hierarchical scene-, motion-group-, and Gaussian-level representation together with a horizon-conditioned transition operator that supports bot…

  • An Accurate and Interpretable Hyper Graph Neural Network for GBM Survival Prediction

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25088v1 Announce Type: new.
      • Abstract: Survival prediction for glioblastoma multiforme (GBM) demands models that are both accurate and interpretable, yet existing approaches treat these objectives as com- peting, where performant models sacrifice transparency, while interpretable models accept degraded predictive power.
      • We argue that this trade-off is not inherent.
      • Graph neural net- works offer a structural foundation for extracting interpretable, explainable representations without compromising discriminative ability.
    • EN Highlights:
      • arXiv:2609.25088v1 Announce Type: new
      • Abstract: Survival prediction for glioblastoma multiforme (GBM) demands models that are both accurate and interpretable, yet existing approaches treat these obj…
      • We argue that this trade-off is not inherent
      • Graph neural net- works offer a structural foundation for extracting interpretable, explainable representations without compromising discriminative ability

  • Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25165v1 Announce Type: new.
      • Abstract: In this report, we introduce \textbf{Ovis-Embedding}, a state-of-the-art series of omni-modal embeddings built on the native integration of text, image, video, and audio.
      • Instead of assembling separate modality towers, Ovis-Embedding encodes different modalities in a common representation space using a shared multi-modal backbone.

  • Specifically, we have made \textbf{three key advances}: (1) \textbf{native omni-modal initialization}: we adopt the pre-trained Qwen-omni model as the embedding backbone and adapt it through low-rank initialized contrastive training; (2) \textbf{data-centric omni-modal training}: we have built an extensive, high-quality corpus covering text, images, videos, audio, and interleaved multimodal data.

    • EN Key Points:
      • arXiv:2609.25165v1 Announce Type: new
      • Abstract: In this report, we introduce \textbf{Ovis-Embedding}, a state-of-the-art omni-modal embedding family built on native integration of text, image, video…
      • Instead of assembling separate modality towers, Ovis-Embedding uses a shared multimodal backbone to encode different modalities in a common representation space
      • Specifically, we make \textbf{three key advances}: (1) \textbf{native omni-modal initialization}: we adopt a pretrained Qwen-omni model as the embedding backbon…

  • X-Planner: Event-Structured Task Planning for Embodied Intelligence

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25187v1 Announce Type: new.
      • Abstract: Task planning bridges high-level instructions and executable behavior in long-horizon manipulation, yet modern Vision-Language-Action (VLA) systems often leave this intermediate structure implicit.
      • Existing chain-of-thought (CoT) planners also tend to rely on coarse task-level annotations or serialize long reasoning traces token by token.
      • We present X-Planner, a planning front-end that addresses both the supervision and representation of embodied reasoning.
    • EN Key Points:
      • arXiv:2609.25187v1 Announce Type: new
      • Abstract: Task planning bridges high-level instructions and executable behavior in long-horizon manipulation, yet modern Vision-Language-Action (VLA) systems of…
      • Existing chain-of-thought (CoT) planners also tend to rely on coarse task-level annotations or serialize long reasoning traces token by token
      • We present X-Planner, a planning front-end that addresses both the supervision and representation of embodied reasoning

  • Lean Pool: An AI-Maintained Archive of Formalized Mathematics

  • Publish Date: 2026-09-23 12:00 Beijing Time

    • Abstract: arXiv:2609.25199v1 Announce Type: new.
      • Abstract: Lean Pool is a repository of formalized mathematics.
      • It is grown, maintained and optimized by AI agents.
      • arXiv:2609.25199v1 Announce Type: new Abstract: Lean Pool is a repository of formalized mathematics It is grown, maintained and optimized by AI agents.
    • EN 要点:
      • arXiv:2609.25199v1 Announce Type: new
      • Abstract: Lean Pool is a repository of formalized mathematics
      • It is grown, maintained and optimized by AI agents.

  • The AI Neuroscientist: An Interactive Agentic Interface for Neuroimaging Analysis

    • Publish Date: 2026-09-23 12:00 Beijing Time
    • Abstract: arXiv:2609.25254v1 Announce Type: new.
      • Abstract: Analyzing neuroimaging data requires specialized coding and statistical expertise, which limits accessibility for researchers without computational backgrounds.
      • We present the AI Neuroscientist, a language agent for interactive data exploration.
      • The system integrates a large language model (LLM) with a neuroimaging toolset to perform quality control, modeling, and visualization.
    • EN 要点:
      • arXiv:2609.25254v1 Announce Type: new
      • Abstract: Analyzing neuroimaging data requires specialized coding and statistical expertise, which limits accessibility for researchers without computational ba…
      • We present the AI Neuroscientist, a language agent for interactive data exploration
      • The system integrates a large language model (LLM) with a neuroimaging toolset to perform quality control, modeling, and visualization

  • MedGate-Fusion: Integrating First-Encounter Semantic Narratives and Physiological Biomarkers for Prospective Stroke Risk Stratification

    • Publish Date: 2026-09-23 12:00 Beijing Time
    • Abstract: arXiv:2609.25272v1 Announce Type: new.

  • Abstract: Prospective stroke risk stratification in primary care is challenging because early risk signals are distributed across routine biomarkers and unstructured clinical narratives.

    • We propose MedGate-Fusion, a multi-modal gated architecture that integrates transformer-based embeddings of first-encounter narratives with ten routinely recorded risk markers.
    • We used electronic medical record data from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN).
    • EN Highlights:
      • arXiv:2609.25272v1 Announce Type: new
      • Abstract: Prospective stroke risk stratification in primary care is challenging because early risk signals are distributed across routine biomarkers and unstruc…
      • We propose MedGate-Fusion, a multi-modal gated architecture that integrates transformer-based embeddings of first-encounter narratives with ten routinely record…
      • We used electronic medical record data from the Canadian Primary Care Sentinel Surveillance Network (CPCSSN)

  • When LLM Agents Fail to Read the Room: ReAdapt for Relational Social Reasoning

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25284v1 Announce Type: new.
      • Abstract: A social agent’s most basic decisions (Should I react to this post?
      • Who should I contact?) are not purely content-related issues.
      • The correct course of action often depends on the underlying relationships between individuals—such as connection strength, reciprocity, and mutual ties—rather than which content is most salient.
    • EN Highlights:
      • arXiv:2609.25284v1 Announce Type: new
      • Abstract: A social agent’s most basic decisions (should I react to this post
      • who should I reach out to
      • ) are not purely content problems

ArXiv cs.CL (B_intro+search) Link to heading

  • What Does 99% Accuracy Measure? A Reproducible Audit of Shortcut Learning in a Widely Used Fake News Corpus

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: [Awaiting Translation] - arXiv:2609.25006v1 Announce Type: new.
      • Abstract: Text classifiers trained on the ISOT/Kaggle “Fake and Real News” corpus routinely report accuracy and F1 above 0.98, a level of performance that sits uneasily beside the difficulty of assessing veracity.

  • Using a transparent TF-IDF and linear-classifier pipeline as a measurement instrument, we audit the corpus along three leakage channels and two distribution-shift protocols, releasing all code and derived numbers.

  • First, the benchmark is partly degenerate: a classifier given only the subject metadata field, with the article text discarded, attains F1 = 1.000, since the two classes have disjoint subjects.

  • EN Highlights:

    • arXiv:2609.25006v1 Announce Type: new
    • Abstract: Text classifiers trained on the ISOT/Kaggle “Fake and Real News” corpus routinely report accuracy and F1 above 0.98, a level of performance that sits…
    • Using a transparent TF-IDF and linear-classifier pipeline as a measurement instrument, we audit the corpus along three leakage channels and two distribution-shi…
    • First, the benchmark is partly degenerate: a classifier given only the subject metadata field, with the article text discarded, attains F1 = 1.000, since the tw…

  • Training a Language Model End-to-End in Rust: An Experience Report

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Summary: [Translation Pending] - arXiv:2609.25008v1 Announce Type: new.
      • Abstract: I pretrained a language model end-to-end in Rust - alone, with no team, no PyTorch, and no Python in the training path - for $164 in rented GPU time.
      • I report that as an achievement, not a recommendation: the more useful contribution is a measured failure taxonomy of the two leading Rust ML frameworks, Candle and Burn, as training (not inference) backends in 2026.
      • I document five Candle defects, including fused kernels that silently produce no gradient, and three Burn defects, including a backward pass at roughly 3% of theoretical GPU throughput and a kernel-fusion path that segfaults mid-training at multi-billion-parameter scale.
    • EN Highlights:
      • arXiv:2609.25008v1 Announce Type: new

  • Abstract: I pretrained a language model end-to-end in Rust - alone, with no team, no PyTorch, and no Python in the training path - for $164 in rented GPU time

  • I report that as an achievement, not a recommendation: the more useful contribution is a measured failure taxonomy of the two leading Rust ML frameworks, Candle…

  • I document five Candle defects, including fused kernels that silently produce no gradient, and three Burn defects, including a backward pass at roughly 3% of th…

  • Same Quantity, Different Answer: Numerical Representation Invariance in Language Models

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25009v1 Announcement Type: new.
      • Abstract: Numerically equivalent word problems should yield the same canonical answer, regardless of whether a quantity is expressed as a decimal, fraction, percentage, number word, scientific notation, or a precisely converted unit.
      • We generated 3,600 exact rational problems and 8,600 prompts, covering five identity-preserving transformation series, and evaluated five open-weight systems.
      • After a standardized, fixed-syntax audit of common answer formats without an LLM judge, the canonical accuracy is 0.969-0.996, but orbit correctness drops to 0.848-0.981, and orbit invariance drops to 0.851-0.981; invariant but incorrect orbits account for at most 0.003.
    • EN Key Points:
      • arXiv:2609.25009v1 Announce Type: new
      • Abstract: Numerically equivalent word problems should yield the same canonical answer whether a quantity is written as a decimal, fraction, percentage, number w…
      • We generate 3,600 exact-rational problems and 8,600 prompts spanning five identity-preserving transformation families, and evaluate five open-weight systems
      • After a fixed syntax audit that normalizes common answer forms without an LLM judge, canonical accuracy is 0.969-0.996, but orbit correctness falls to 0.848-0.9…

  • A Computational Approach to Measuring Semantic Change in Sanskrit Literature

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25012v1 Announce Type: new.
      • Abstract: Diachronic word embeddings have become the modern standard for tracking semantic change, yet they have been largely validated on modern, high-resource, and well-segmented languages.

  • This paper tests whether the paradigm transfers to Sanskrit, an ancient, low-resource language whose phonological fusion (sandhi), morphological inflection, compounding, and polysemy pose a unique challenge.

    • I assemble a 2.7M-token corpus spanning four canonical periods, recover word boundaries with a neural byte-level sandhi splitter and lemmatizer, and train per-period embeddings across configurations.
    • EN Key Points:
      • arXiv:2609.25012v1 Announce Type: new
      • Abstract: Diachronic word embeddings have become the modern standard for tracking semantic change, yet they have been largely validated on modern, high-resource…
      • This paper tests whether the paradigm transfers to Sanskrit, an ancient, low-resource language whose phonological fusion (sandhi), morphological inflection, com…
      • I assemble a 2.7M-token corpus spanning four canonical periods, recover word boundaries with a neural byte-level sandhi splitter and lemmatizer, and train per-p…

  • Retrieved-Span Training for Efficient Query-Focused Meeting Summarization on QMSum

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25028v1 Announce Type: new.
      • Abstract: QMSum provides no scorer, making query-focused meeting summarization results difficult to compare.
      • We rescore or generate 15 systems under one implementation.
      • Through a common inference port, a released 406M Fusion-in-Decoder specialist loses 6.30 ROUGE-1 when moved from capped long input to 2,000-word retrieved spans.
    • EN Key Points:
      • arXiv:2609.25028v1 Announce Type: new
      • Abstract: QMSum provides no scorer, making query-focused meeting summarization results difficult to compare
      • We rescore or generate 15 systems under one implementation
      • Through a common inference port, a released 406M Fusion-in-Decoder specialist loses 6.30 ROUGE-1 when moved from capped long input to 2,000-word retrieved spans

  • From Tone to Trajectory: Continuous Sentiment and the Shape of Monetary Policy Communication

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25034v1 Announce Type: new.
      • Abstract: Central bank press conferences are not just information releases—they are structured narratives.
      • We study whether the shape of sentiment in the statement (and not just its average tone) carries policy-relevant signals.

  • By constructing sentiment arcs for the press conferences of the European Central Bank and the Federal Reserve along three dimensions—monetary stance, economic outlook, and uncertainty—we assess their predictive content on policy rate changes, inflation expectations, and forecast disagreement.

    • EN Highlights:
      • arXiv:2609.25034v1 Announce Type: new
      • Abstract: Central bank press conferences are not merely information releases — they are structured narratives
      • We study whether the shape of sentiment within a statement, not just its average tone, carries policy-relevant signals
      • Constructing sentiment arcs for ECB and Fed press conferences along three dimensions — monetary stance, economic outlook, and uncertainty — we assess their…

  • Peerify: Benchmarking Peer-Review Claim Verification

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25046v1 Announce Type: new.
      • Abstract: Peer review plays a central role in scholarly publishing, yet verifying whether reviewer claims are supported by manuscript evidence remains a largely manual and time-consuming process.
      • We present Peerify, a pipeline for manuscript-grounded verification of peer-review claims.
      • Given a manuscript and a review comment, the Peerify pipeline decomposes reviews into atomic claims, retrieves relevant manuscript evidence, and determines whether each claim is supported by the paper.
    • EN Highlights:
      • arXiv:2609.25046v1 Announce Type: new
      • Abstract: Peer review plays a central role in scholarly publishing, yet verifying whether reviewer claims are supported by manuscript evidence remains a largely…
      • We present Peerify, a pipeline for manuscript-grounded verification of peer-review claims
      • Given a manuscript and a review comment, the Peerify pipeline decomposes reviews into atomic claims, retrieves relevant manuscript evidence, and determines whet…

  • AIBuildAI-2.5: Efficient Autonomous AI Model Development Through LLM-Guided Tree Search

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25047v1 Announce Type: new.

  • Abstract: Autonomous agents that automatically build artificial intelligence (AI) models could broaden access to AI across science and engineering.

  • A popular line of such agents frames model building as a code search problem and solves it by tree search, in which each node is a candidate program and the tree grows by generating a child program from a parent, and these agents now approach the capability of experienced AI engineers on realistic benchmarks.

  • However, these agents have three weaknesses in efficiency that have not been fully addressed.

  • EN Highlights:

    • arXiv:2609.25047v1 Announce Type: new
    • Abstract: Autonomous agents that automatically build artificial intelligence (AI) models could broaden access to AI across science and engineering
    • A popular line of such agents frames model building as a code search problem and solves it by tree search, in which each node is a candidate program and the tre…
    • However, these agents have three weaknesses in efficiency that have not been fully addressed

  • Prompt Breadth and Rollout Refresh Interact in On-Policy Distillation

    • Published: 2026-09-23 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.25048v1 Announce Type: new.
      • Abstract: How many prompts does on-policy distillation (OPD) need, and how does the answer depend on the student policies that generate its training responses?
      • We study these two controls jointly: prompt breadth and rollout refresh.
      • A 3x3 mathematical-reasoning experiment fixes 14,080 trajectories and 110 optimizer updates while varying the prompt bank and the number of response-generating policy snapshots.
    • EN Highlights:
      • arXiv:2609.25048v1 Announce Type: new
      • Abstract: How many prompts does on-policy distillation (OPD) need, and how does the answer depend on the student policies that generate its training responses
      • We study these two controls jointly: prompt breadth and rollout refresh

  • A 3x3 mathematical-reasoning experiment fixes 14,080 trajectories and 110 optimizer updates while varying the prompt bank and the number of response-generating…

  • Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibratione

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25049v1 Announce Type: new.
      • Abstract: Large language models (LLMs) aligned for safety often suffer from over-refusal, incorrectly rejecting benign yet safety-related instructions.
      • Prior studies primarily attribute this to static representation overlap, largely overlooking the underlying dynamic mechanisms.
      • In this paper, we present the mechanistic analysis of over-refusal through the lens of internal routing conflicts within transformer attention.
    • EN Highlights:
      • arXiv:2609.25049v1 Announce Type: new
      • Abstract: Large language models (LLMs) aligned for safety often suffer from over-refusal, incorrectly rejecting benign yet safety-related instructions
      • Prior studies primarily attribute this to static representation overlap, largely overlooking the underlying dynamic mechanisms
      • In this paper, we present the mechanistic analysis of over-refusal through the lens of internal routing conflicts within transformer attention

ArXiv cs.LG (B_intro+search) Link to heading

  • “As a Language Model…”: Chat Template Switches LLM Self-Referential Voice and Activation Steering Reproduces It

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25021v1 Announce Type: new.
      • Abstract: Large Language Models (LLMs) tend to add disclaimers like “I’m just an AI” when asked about something related to themselves.
      • The self-reports from such responses are used in debates about AI safety or self-knowledge of the models, yet what drives them is not well understood.
      • Are the models telling us about themselves or rather how they are deployed?
    • EN Highlights:
      • arXiv:2609.25021v1 Announce Type: new

  • Abstract: Large Language Models (LLMs) tend to add disclaimers like “I’m just an AI” when asked about something related to themselves

  • The self-reports from such responses are used in debates about AI safety or self-knowledge of the models, yet what drives them is not well understood

  • Are the models telling us about themselves or rather how they are deployed

  • Federating Quantum and Classical Computing: A Privacy-Preserving Hybrid Approach

    • Posted: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25082v1 Announcement Type: new.
      • Abstract: Quantum Machine Learning (QML) is increasingly recognized as one of the most promising near-term applications of quantum computing and is seen as the next frontier candidate to surpass purely classical methods.
      • Hybrid quantum-classical models harness this potential by embedding parameterized quantum circuits into the model, where all other components remain classical—a design that has been applied to chemical simulation, financial modeling, and image classification.
      • However, their deployment in privacy-sensitive, multi-party environments is limited by the need to avoid centralizing raw data and the requirement that modern quantum circuits remain parameter-efficient to stay scalable for training.
    • EN Highlights:
      • arXiv:2609.25082v1 Announce Type: new
      • Abstract: Quantum machine learning (QML) is increasingly recognized as one of the most promising near-term applications of quantum computing, viewed as a next-f…
      • Hybrid quantum-classical models operationalize this potential by embedding a parameterized quantum circuit within a model where all other components remain clas…
      • However, their deployment in privacy-sensitive, multi-party settings is constrained by the need to avoid centralizing raw data and by the requirement that moder…

  • Entropy Can Flow, or It Can Guide. Be Entropy. LEDFlow: Introducing Entropy-guided Generation Order into Uniform Discrete Flow

    • Posted: 2026-09-23 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.25131v1 Announce Type: new.
      • Abstract: Uniform discrete flow permits repeated updates at every generation position.
      • While continued revision supports correction of wrong tokens, it also exposes correct intermediate predictions to later errors.
      • An experiment on Sudoku puzzles shows that 9.4% of generated cells are correct at an intermediate step but incorrect in the final output.
    • EN Highlights:
      • arXiv:2609.25131v1 Announce Type: new

  • Abstract: Uniform discrete flow permits repeated updates at every generation position

    • While continued revision supports correction of wrong tokens, it also exposes correct intermediate predictions to later errors
    • An experiment on Sudoku puzzles shows that 9.4% of generated cells are correct at an intermediate step but incorrect in the final output

  • The Probabilistic Structure of Large Language Models

    • Published: 2026-09-23 12:00 Beijing Time
    • Summary: [To be translated] - arXiv:2609.25134v1 Announce Type: new.
      • Abstract: This paper presents a probabilistic perspective on large language models (LLMs), developed with the aim of bringing together, in a single self-contained account, tools that are usually treated separately across the literature.
      • LLMs are described through probability measures on the set of sequences of tokens, specified via their autoregressive conditional distributions.
      • Training is formulated as a maximum-likelihood estimation problem, addressed by stochastic gradient methods, while text generation is viewed as the sequential simulation of the resulting stochastic process.
    • EN Highlights:
      • arXiv:2609.25134v1 Announce Type: new
      • Abstract: This paper presents a probabilistic perspective on large language models (LLMs), developed with the aim of bringing together, in a single self-contain…
      • LLMs are described through probability measures on the set of sequences of tokens, specified via their autoregressive conditional distributions
      • Training is formulated as a maximum-likelihood estimation problem, addressed by stochastic gradient methods, while text generation is viewed as the sequential s…

  • Stable Unsupervised Continual Chunking with Sheaf SyncMap

    • Published: 2026-09-23 12:00 Beijing Time
    • Summary: [To be translated] - arXiv:2609.25143v1 Announce Type: new.

  • Abstract: Unsupervised Continual chunking is a fundamental problem in machine learning and neuroscience, where the goal is to identify groups of states that frequently co-occur in temporal sequences.

    • A key challenge is to form accurate chunks while maintaining their stability over time.
    • In this work, we propose sheaf regularization to reduce local inconsistencies in Decentralized SyncMap, a self-organizing system, and thereby stabilize its chunking dynamics.
    • EN Key Points:
      • arXiv:2609.25143v1 Announce Type: new
      • Abstract: Unsupervised Continual chunking is a fundamental problem in machine learning and neuroscience, where the goal is to identify groups of states that fre…
      • A key challenge is to form accurate chunks while maintaining their stability over time
      • In this work, we propose sheaf regularization to reduce local inconsistencies in Decentralized SyncMap, a self-organizing system, and thereby stabilize its chun…

  • Brain-Inspired Hierarchical Modularity for General Continual Learning

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.25146v1 Announce Type: new.
      • Abstract: Continual learning, the ability to learn from sequential experience while retaining and adapting prior knowledge, is central to intelligent systems operating in changing environments.
      • However, conventional continual learning is typically studied with offline task-wise training and clear task boundaries, leaving a substantial gap from general continual learning under online, uncertain, and evolving data streams.
      • In this regime, intelligent systems must separate conflicting experience to reduce interference while integrating compatible experience to promote generalization.
    • EN Key Points:
      • arXiv:2609.25146v1 Announce Type: new
      • Abstract: Continual learning, the ability to learn from sequential experience while retaining and adapting prior knowledge, is central to intelligent systems op…

  • However, conventional continual learning is typically studied with offline task-wise training and clear task boundaries, leaving a substantial gap from general…

  • In this regime, intelligent systems must separate conflicting experience to reduce interference while integrating compatible experience to promote generalizatio…

  • Dual-GNN Multilevel Coarsening for Maximum Independent Set

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Summary: - arXiv:2609.25149v1 Announcement Type: New.
      • Abstract: Solving large-scale instances of the Traveling Salesman Problem (TSP) is computationally expensive.
      • Researchers often employ graph sparsification methods to improve computational efficiency.
      • Traditional sparsification methods typically rely on fixed heuristics and fail to fully exploit instance-specific structural information.
    • EN Highlights:
      • arXiv:2609.25149v1 Announce Type: new
      • Abstract: Solving large-scale instances of the Traveling Salesman Problem (TSP) exactly is computationally expensive
      • Researchers often employ graph sparsification methods to improve computational efficiency
      • Traditional sparsification methods typically rely on fixed heuristics and fail to fully exploit instance-specific structural information

  • Exposing Blind Spots in Deep Imbalanced Regression Evaluation

    • Release Time: 2026-09-23 12:00 Beijing Time
    • Summary: - arXiv:2609.25152v1 Announcement Type: New.
      • Abstract: Deep Imbalanced Regression (DIR) addresses a common failure mode of regression models: the target distribution is highly non-uniform, causing models to perform best in densely populated target regions, even when reliable performance is needed across the entire target range.
      • Despite rapid methodological progress, DIR evaluation remains constrained by three blind spots: it is dominated by image-based benchmarks, its standard many-/medium-/few-shot protocol is diagnostic but not decision-complete, and the tail-region stability of random seeds has not been systematically evaluated.
      • We revisit DIR evaluation along these three axes.
    • EN Highlights:
      • arXiv:2609.25152v1 Announce Type: new
      • Abstract: Deep Imbalanced Regression (DIR) addresses a common failure mode of regression models: target distributions are highly non-uniform, causing models to…
      • Despite rapid methodological progress, DIR evaluation remains constrained by three blind spots: it is dominated by image-based benchmarks, its standard many-/me…
      • We revisit DIR evaluation along these three axes

  • Learning Neural Feedback Linearization for Data-driven Systems via Augmented Lagrangian

    • Release Time: 2026-09-23 12:00 Beijing Time

  • Abstract: - arXiv:2609.25163v1 Announce Type: new.

    • Abstract: This paper proposes a novel data-driven framework for designing and training a feedback linearizing controller by explicitly incorporating conditions based on relative degree into the learning process.
    • This enables the conventional feedback controller components to be replaced by neural Lie derivatives, thereby facilitating a fully data-driven feedback linearization framework.
    • Furthermore, practical closed-loop stability is established by deriving sufficient conditions under which bounded identification errors lead to bounded tracking errors.

  • EN Key Points:

    • arXiv:2609.25163v1 Announce Type: new
    • Abstract: The paper proposes a novel data-driven framework for designing and training a feedback linearizing controller by explicitly incorporating relative deg…
    • This enables the conventional feedback controller components to be replaced by neural Lie derivatives, thereby facilitating a fully data-driven feedback lineari…
    • Furthermore, practical closed-loop stability is established by deriving sufficient conditions under which bounded identification errors lead to bounded tracking…

  • Mitigating Sequential Reappearance in Diffusion Data-Point Unlearning

    • Publication Time: 2026-09-23 12:00 Beijing Time
    • Abstract: - arXiv:2609.25166v1 Announce Type: new.
      • Abstract: Diffusion data-point unlearning is typically evaluated immediately after each deletion, even though subsequent requests may repeatedly update the same model.
      • We identify sequential reappearance, a failure mode in which an instance that is initially judged to be forgotten later returns to the memorized regime without reusing the deleted data or undergoing adversarial fine-tuning.
      • To capture this behavior, we introduce a target-level evaluation protocol that tracks whether each target is forgotten immediately, remains forgotten at the end of the sequence, or reappears during subsequent deletions.
    • EN Key Points:
      • arXiv:2609.25166v1 Announce Type: new
      • Abstract: Diffusion data-point unlearning is typically evaluated immediately after each deletion, even though subsequent requests may repeatedly update the same…
      • We identify sequential reappearance, a failure mode in which an instance that is initially judged to be forgotten later returns to the memorized regime without…
      • To capture this behavior, we introduce a target-level evaluation protocol that tracks whether each target is forgotten immediately, remains forgotten at the end…