System translated (Gemini)

🤖 AI 速览

Today, the application boundaries of AI continue to expand: On one hand, specialized products connecting enterprise data and serving financial research are taking shape; on the other hand, Codex and ChatGPT have been used to search for antimicrobial molecules, driving AI from information processing …
📋 文章元数据
发布时间
2026-09-11
类型
ai-daily
字数
7368
阅读时长
35 min

2026-09-11 AI Daily | From Financial Research to Antimicrobial Drug Discovery, AI Accelerates Entry into High-Value Decision Chains Link to heading

Today, the boundaries of AI application continue to expand: on one hand, specialized products are emerging to connect enterprise data and serve financial research; on the other, Codex and ChatGPT are being used to search for antimicrobial molecules, advancing AI from information processing to scientific discovery. Meanwhile, research traceability, result reliability, and data permissions are becoming fundamental issues that must be addressed for large-scale adoption.

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

The most noteworthy trend today is AI’s accelerating transition from a “general-purpose assistant” to high-value workflows: OpenAI has launched ChatGPT for financial services and expanded its support for government cybersecurity. Simultaneously, researchers are using Codex and ChatGPT to search for antimicrobial molecules, showing that AI’s value is shifting from Q&A to research and decision-making. The second key theme is reliability and auditability: several papers examine model performance in cross-lingual factual refusal, handling erroneous data, in-depth research, and medical triage, highlighting how AI is being validated in real-world scenarios. The third theme focuses on infrastructure, where X-CoSD and Osprey explore lower-cost inference acceleration through collaborative speculative decoding and draft model training, respectively.

🌐 AI Hot Topics on X Link to heading

Topic 1:DeepSeek Launches V4.1 Flash: Faster AI at Lower Cost Link to heading

  • Category: AI · News
  • Overview: Trending for: 2 days ago, Related posts: 47,000
  • What it is: DeepSeek released V4.1 Flash, featuring faster response speeds and lower usage costs.
  • Why it matters: Higher inference efficiency and lower costs could expand the applications of generative AI and intensify competition in the model-as-a-service market.
  • Discussion summary: Discussions on X focus on its actual speed and cost advantages, whether model performance remains stable, and its competitive performance against other cost-effective AI models.

Topic 2:OpenAI Launches Data Agent for ChatGPT Work Link to heading

  • Category: AI · News
  • Overview: Trending for: 8 hours ago, Related posts: 1,900
  • What it is: OpenAI is rolling out Data Agent to ChatGPT Work users, enabling connections to various enterprise data sources to answer analytical questions, generate dashboards, and execute actions across tools.
  • Why it matters: This marks a shift for enterprise-grade AI assistants from content generation to directly accessing and manipulating distributed business data, potentially reducing data integration costs and promoting the adoption of agent-based analytics.
  • Discussion summary: Discussions on X are centered on the range of supported data sources, the convenience of not needing centralized data migration, and concerns about enterprise data permissions, security, and operational risks. Some users question OpenAI’s failure to release benchmarks simultaneously, making it difficult to independently assess its analytical accuracy and real-world performance.

Topic 3:OpenAI Launches ChatGPT for Financial Services Targeting Bankers Link to heading

  • Category: AI · News
  • Overview: Trending for: 5 hours ago, Related posts: 4,300
  • What it is: OpenAI has launched a version of ChatGPT for the financial services industry, primarily serving investment bankers and equity researchers in their research, modeling, and client material preparation tasks.
  • Why it matters: This indicates that generative AI is moving from a general-purpose assistant into high-value, heavily regulated professional financial scenarios, and intensifies the competition between OpenAI and Anthropic for enterprise clients.
  • Discussion summary: Discussions on X are focused on whether AI will replace junior investment banking staff, the accuracy and traceability of financial data and model outputs, regulatory and confidentiality risks, and the competition between OpenAI and Anthropic in the financial enterprise market.

Topic 4:OpenAI Pauses New $200 ChatGPT Pro Sign-Ups Amid Astra Demand Surge Link to heading

  • Category: AI · News
  • Overview: Trending for: 1 day ago, Related posts: 9,500
  • What it is: OpenAI has paused new sign-ups for its $200/month ChatGPT Pro due to a surge in demand for a new feature codenamed “Astra.”
  • Why it matters: This reflects that demand growth for high-end AI subscription services may be outpacing compute and service capacity. It also highlights the tension between commercialization, resource allocation, and user base control for cutting-edge AI products.
  • Discussion summary: Discussions on X center on the specific capabilities and appeal of “Astra,” whether the pause in sign-ups is due to a compute shortage or a marketing strategy, the reasonableness of the $200 price point, and OpenAI’s ability to reliably provide high-end services.

Topic 5:Job Interview Meme Envisions Robot Gunfights in Five Years Link to heading

  • Category: AI · Entertainment
  • Overview: Trending for: 1 day ago, Related posts: 15,000
  • What it is: A job interview meme about “robot gunfights happening in five years” has gone viral on X.
  • Why it matters: This topic, in an exaggerated and entertaining way, reflects public concern about the rapid development of robotics, the militarization of AI, and future security risks.
  • Discussion Overview: The discussion primarily revolves around whether this is just dark humor, whether robots and autonomous weapons will accelerate into reality, and how AI development should be regulated, with supporters and skeptics divided on its realistic possibility.

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

The main theme of today’s discourse is that AI is accelerating its implementation along two paths simultaneously: “lower cost, higher efficiency” and “deep integration into enterprise and professional scenarios.” From model services and data analysis to financial research, commercial competition is clearly heating up. There is a general consensus on the expanding scope of AI applications, the improvement in corporate productivity, and the growing demand for premium subscriptions. However, the actual performance of DeepSeek V4.1 Flash, the analytical accuracy of OpenAI’s Data Agent, and the substitution effect in financial scenarios still lack sufficient validation. Disagreements are mainly centered on whether the suspension of registrations for high-priced products is due to insufficient computing power or a marketing strategy, the extent to which AI will replace junior professionals, and whether the risk of robot militarization is a real threat or an exaggeration. Potential risks include data permissions and privacy leaks, errors in financial decision-making and difficulties in liability tracing, insufficient service stability, and the security and regulatory pressures brought by autonomous systems and the proliferation of AI in military applications.

💡 Influencer Insights Link to heading

2026.09.10 AI Industry Daily Briefing Link to heading

1. Key Trend: From Vibe Coding to the Deep Waters of “Model-as-a-Product” Link to heading

The focus of today’s discussion among influencers has shifted from simply climbing model leaderboards to the fundamental transformation of AI programming paradigms and the deep game of model commercialization.

  • Vibe Coding is entering the “tooling” and “long-tail” phases: @dotey (Baoyu), through his observations on Weibo VibeLab, points out that the barrier to entry for software development has been lowered from “professional skills” to “language expression.” The core trend is no longer about replacing programmers, but about empowering ordinary people to “build tools” to meet their own long-tail needs. He specifically mentioned Shopify’s case of abandoning React Native to return to native development, revealing how AI breaks the boundaries of tech stacks and empowers developers to restructure their productivity.
  • The “Consumer-Grade Implementation” and “Performance Degradation Mystery” of Model Capabilities: @Pluvio9yte and @vista8 both independently reported fluctuations in the performance of GPT-6 Astra within Codex. Although the model has a very high capability ceiling, phenomena of “severe performance degradation” and abnormal Token quota displays occur frequently. This has led to speculation about the stability of cloud-based models and whether service providers are controlling costs through variable kernels (e.g., by dynamically reducing inference precision).
  • The Traffic Logic of “Model as the Object of Payment”: @gefei55 (Gefei) put forward a sharp and provocative theory: users don’t pay for AI SaaS, they pay for the underlying AI model. Combined with the case he shared about using ChatGPT to drive traffic to a product, it shows that in the fields of SEO and international expansion, bundling top-tier model capabilities and skillfully using AI search optimization is key to acquiring high-value customers.

2. Unique Perspectives and Industry Foresight Link to heading

  • On-Device AI and the Tipping Point of “Proactive Memory”: @zhixianio (Zhixian) shared the “‘Sycophant Module’” he is working on for personal AI scenarios. This signals that personal AI is evolving from a passive Q&A machine into an intelligent agent that understands the user’s environment in real-time through on-device model clusters and proactively feeds them information. He believes that running multi-model scheduling on consumer-grade devices is the next key battleground.
  • The “Red Queen Effect” in Model Competition: @lijigang (Li Jigang) proposed that model vendors are caught in a “Red Queen Effect” game where “falling behind means death.” Even if the market feels fatigued, the entire industry is forced to run at an extremely fast pace of iteration, and any pause could mean being eliminated from the game.
  • AI Reshaping Interaction and Design Philosophy: @nishuang (Ni Shuang) and @lijigang both mentioned the shift in the focus of interaction design. Ni Shuang believes that design in the AI era should abandon complex gamified rewards and return to a “sense of play,” allowing users to consume content for enjoyment rather than task completion. Li Jigang, on the other hand, emphasizes that interaction should “move to the user’s dance steps.”

Hardcore Deployment and Development

  • Obsidian AI Plugin Suite: @vista8 developed an RSS subscription, an AI reading companion (all formats), and a general AI assistant plugin using GPT-6 Astra, suitable for users focused on knowledge management.
  • Flatrouter: A model routing platform recommended by @Pluvio9yte, featuring extremely low prices (GPT Image 2.5 for only 0.05 CNY/image), suitable for users creating image generation applications or conducting batch testing.
  • OpenConnector (Open Password Gateway): An open-source tool recommended by @ruanyf that solves the pain point of AI Agent password leakage by centrally managing connection authorizations without exposing credentials.

Content and Efficiency

  • GPT-Image 2.5 Resource Compilation: @Pluvio9yte has compiled over 25 tested viral posts, including Sketch-to-image conversion, multi-turn refinement, and transparent background generation. It is currently the most comprehensive library of first-hand experiences with the new image model.
  • Doubao Desktop Work Suite: @Pluvio9yte and @vista8 are both recommending it. For ordinary users in China with a non-technical background, the maturity of Doubao PC version and Input Method is sufficient to meet daily AI intervention needs, and there are no network barriers.
  • US Region Communications and Payments: @AI_Jasonyu published “2026 US Region PayPal Business Account Opening Guide” and T-Mobile ENC Card Activation Tutorial, which are fundamental for fully experiencing the complete range of overseas AI services (especially those requiring physical number verification).

Opinions and Thoughts: Beyond tools, it is recommended to pay attention to @dotey’s recommendation on “Illustrated Skill” and @zhixianio’s podcast on “Proactive Memory Systems for Personal AI”, which are underlying mindsets that need to be supplemented for those engaged in AI application layer development.

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

Time Window: Last 3 days; Covering 22 sources; Total 38 updates

Stratechery by Ben Thompson (A_full) Link to heading

  • The iPhone Duo, The Intelligent Personal Hub, Apple Watch Audio Intelligence
    • Publication Time: 2026-09-10 18:00 Beijing Time
    • Summary: Apple once again demonstrated the power of integrating hardware and software, but its biggest AI blindspot might be its belief in the primacy of apps.
      • $15 / month or $150 / year.
      • Substantial analysis of the news of the day delivered via three weekly emails or podcasts.
      • Stratechery Interviews.
      • Interviews with leading public CEOs, private company founders, and discussions with fellow analysts.
    • Key English Points:
      • Apple once again demonstrated the power of integrating hardware and software, but it’s biggest AI blindspot might be its belief in the primacy of apps.

OpenAI Blog (A_full) Link to heading

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

    • Publication Time: 2026-09-11 00:00 Beijing Time
    • Summary: Drug-resistant microbes including bacteria, fungi, parasites, and viruses are a growing global threat.
      • It can take years to find molecules with the potential to become antimicrobials.
      • Researchers are using AI to accelerate this early stage of discovery.
      • “And yet, we haven’t had a new class of antibiotics for 50 years.”.
      • Much of modern antimicrobial development focuses on modifying existing medicines or searching familiar classes of chemicals.
    • Key English Points:

  • César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.

  • Now everyone can put data to work

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

  • Introducing ChatGPT for Financial Services

    • Publication Time: 2026-09-10 15:00 Beijing Time
    • Summary: Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.
      • This piece from OpenAI Blog explains how Introducing ChatGPT for Financial Services shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Introducing ChatGPT for Financial Services.
    • EN Key Points:
      • Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.

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

    • Publication Time: 2026-09-10 15:00 Beijing Time
    • Summary: OpenAI and GSA will offer eligible federal, state, local, and tribal governments $0 license fees, 50% off usage, and expanded cyber defense support.

  • This piece from the OpenAI Blog explains how expanding AI access and cyber defense for federal, state, local, and tribal governments shapes the broader AI and infrastructure landscape.

  • It also surfaces practical implications for founders, operators, and investors following this expansion.

  • EN Key points:

    • OpenAI and GSA will offer eligible federal, state, local, and tribal governments $0 license fees, 50% off usage, and expanded cyber defense support.

  • Introducing the Agents API

    • Release Time: 2026-09-10 08:00 Beijing Time
    • Summary: [Translation pending] - Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.
      • This piece from the OpenAI Blog explains how the introduction of the Agents API shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following the introduction of the Agents API.
    • EN Key points:
      • Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.

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

    • Release Time: 2026-09-10 08:00 Beijing Time
    • Summary: [Translation pending] - GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.
      • This piece from the OpenAI Blog explains how building more natural voice experiences with GPT‑Live‑1 in the API shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following the ability to build more natural voice experiences with GPT‑Live‑1 in the API.
    • EN Key points:
      • GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.

Two Minute Papers (B_intro+search) Link to heading

  • I Never Thought I’d See This Happen
    • Published: 2026-09-10 16:47 Beijing Time
    • Summary: [TO BE TRANSLATED] - ❤️ Check out Lambda here and sign up for their GPU Cloud:.
      • 📝 The Navier-Stokes solution paper is available here:.
      • My fluid simulations and papers:.
      • The flow from simulation to reality:.
    • EN Key Points:
      • ❤️ Check out Lambda here and sign up for their GPU Cloud:
      • 📝 The Navier-Stokes solution paper is available here:
      • My fluid simulations and papers:
      • The flow from simulation to reality:

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

  • OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows

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

  • Adaptive Entangled Game Modules in Artificial General Intelligence

    • Published: 2026-09-10 12:00 Beijing Time

  • Abstract: - arXiv:2609.09226v1 Announce Type: new.

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

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: - arXiv:2609.09233v1 Announce Type: new.
      • Abstract: How can language model agents effectively leverage libraries of reusable knowledge to solve long-horizon tasks?
      • Recent work has increasingly focused on agent skills: reusable capabilities represented as skill packages, i.e., multi-file bundles containing instructions, scripts, and other resources that help agents perform specific tasks.

  • Agent skills are typically executed by loading their skill instructions into an agent’s context and relying on the agent to follow them.

    • EN Key Points:
      • arXiv:2609.09233v1 Announce Type: new
      • Abstract: How can language model agents effectively leverage libraries of reusable knowledge to solve long-horizon tasks
      • Recent work has increasingly focused on agent skills: reusable capabilities represented as skill packages, i.e., multi-file bundles containing instructions, scr…
      • Agent skills are typically executed by loading their skill instructions into an agent’s context and relying on the agent to follow them

  • Gradland: On Phenomenal Experience, Differentiated Across Many Dimensions

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09306v1 Announce Type: new.
      • Abstract: This paper investigates the hypothesis that the first-order structure of physical interactions, i.e.
      • gradients or Jacobians, characterizes the structure of phenomenal experience.
      • It does so in an idealized world inhabited by neural networks, Gradland, where the physics are known and the functions are (mostly) differentiable.
    • EN Key Points:
      • arXiv:2609.09306v1 Announce Type: new
      • Abstract: This paper investigates the hypothesis that the first-order structure of physical interactions, i.e
      • gradients or Jacobians, characterizes the structure of phenomenal experience
      • It does so in an idealized world inhabited by neural networks, Gradland, where the physics are known and the functions are (mostly) differentiable

  • An Autonomous GeoAI Agent for Arctic Eco-Navigation

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09374v1 Announce Type: new.
      • Abstract: Arctic maritime navigation is becoming increasingly important as changing sea-ice conditions expand seasonal accessibility while simultaneously introducing substantial operational, environmental, and community risks.

  • Arctic route planning is inherently a multi-criteria problem: routes that improve vessel safety or efficiency may increase exposure to sea ice, sensitive ecosystems, or nearby communities.

    • Existing routing methods prioritize travel time, fuel use, and navigational risk, often overlooking ecological and community impacts.
    • EN Key Points:
      • arXiv:2609.09374v1 Announce Type: new
      • Abstract: Arctic maritime navigation is becoming increasingly important as changing sea-ice conditions expand seasonal accessibility while simultaneously introd…
      • Arctic route planning is inherently a multi-criteria problem: routes that improve vessel safety or efficiency may increase exposure to sea ice, sensitive ecosys…
      • Existing routing methods prioritize travel time, fuel use, and navigational risk, often overlooking ecological and community impacts

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09395v1 Announce Type: new.
      • Abstract: Language models act through tools, yet practical agents face libraries containing thousands of interfaces.
      • We introduce the tool menu as the short, ordered subset of available tools shown to an agent before execution.
      • The agent can call only tools in this menu.
    • EN Key Points:
      • arXiv:2609.09395v1 Announce Type: new
      • Abstract: Language models act through tools, yet practical agents face libraries containing thousands of interfaces
      • We introduce the tool menu as the short, ordered subset of available tools shown to an agent before execution
      • The agent can call only tools in this menu

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09413v1 Announce Type: new.

  • Abstract: Choosing a recovery process for scale-up requires connecting laboratory results with product requirements, process costs, and scale effects.

    • We analyze records from Pacific Northwest National Laboratory’s Computer Intelligence for Critical Element Recovery and Optimization (CICERO) workflow for autonomous selective precipitation.
    • Active learning uses prior results to choose experiments.
    • EN Highlights:
      • arXiv:2609.09413v1 Announce Type: new
      • Abstract: Choosing a recovery process for scale-up requires connecting laboratory results with product requirements, process costs, and scale effects
      • We analyze records from Pacific Northwest National Laboratory’s Computer Intelligence for Critical Element Recovery and Optimization (CICERO) workflow for auton…
      • Active learning uses prior results to choose experiments

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

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

  • Existing game-oriented approaches often combine action-conditioned world models with external policies and reward functions to realize WAM-like decision-making,…

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

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

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Pending Translation] - arXiv:2609.09448v1 Announce Type: new.
      • Abstract: As agentic systems getting adopted rapidly in safety critical applications, it is vital to measure the confidence associated with the agentic actions.

  • In comparison to the traditional machine learning systems, agentic workflows have complex failure modes with planning, tool invocation and dynamic environment interactions.

    • In this paper, we investigate whether model’s internal representations provide stronger signals of eventual task success in multi-turn agentic setups.
    • EN Key Points:
      • arXiv:2609.09448v1 Announce Type: new
      • Abstract: As agentic systems getting adopted rapidly in safety critical applications, it is vital to measure the confidence associated with the agentic actions
      • In comparison to the traditional machine learning systems, agentic workflows have complex failure modes with planning, tool invocation and dynamic environment i…
      • In this paper, we investigate whether model’s internal representations provide stronger signals of eventual task success in multi-turn agentic setups

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

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation pending] - arXiv:2609.09166v1 Announce Type: new.
      • Abstract: This paper investigates collaborative speculative decoding (CoSD), a distributed large language model (LLM) inference framework in which an on-device small language model (SLM) drafts candidate tokens and a server LLM verifies them.
      • Existing CoSD methods assume a shared vocabulary between the SLM and the LLM and incur substantial communication load because residual resampling requires token distribution exchange between the user device and the edge server.
      • To address these limitations, we propose cross-vocabulary CoSD (X-CoSD), a lossless and communication-efficient CoSD framework for heterogeneous SLM-LLM vocabularies.
    • EN Key Points:
      • arXiv:2609.09166v1 Announce Type: new
      • Abstract: This paper investigates collaborative speculative decoding (CoSD), a distributed large language model (LLM) inference framework in which an on-device…

  • Existing CoSD methods assume a shared vocabulary between the SLM and the LLM and incur substantial communication load because residual resampling requires token…

  • To address these limitations, we propose cross-vocabulary CoSD (X-CoSD), a lossless and communication-efficient CoSD framework for heterogeneous SLM-LLM vocabul…

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

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

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

    • Publication Time: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09338v1 Announce Type: new.

  • Abstract: Speculative decoding is critical for accelerating LLM inference.

  • However, the speedup is fragile: drafters are typically trained against a narrow distribution for a single target model, and their acceptance rate collapses under workload shifts.

  • This is a striking inversion of modern LLM development, where target models are valued precisely for the broad generalization they acquire through large-scale pretraining.

    • EN Highlights:
      • arXiv:2609.09338v1 Announce Type: new
      • Abstract: Speculative decoding is critical for accelerating LLM inference
      • However, the speedup is fragile: drafters are typically trained against a narrow distribution for a single target model, and their acceptance rate collapses und…
      • This is a striking inversion of modern LLM development, where target models are valued precisely for the broad generalization they acquire through large-scale p…

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

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

  • We introduce Systematic Wikidata-based Object-Relation Distortion (SWORD), a benchmark that evaluates whether models consistently reject factual errors across l…

  • SWORD generates syntactically well-formed but factually incorrect statements in eight widely spoken languages through controlled perturbations of Wikidata tripl…

  • Auditable Emergency Triage for Maternal and Newborn Care in India

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

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09363v1 Announce Type: new.
      • Abstract: Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or data-to-text systems.
      • We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict).

  • To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on local knowledge, only partially captured in models’ parametric knowledge.

    • EN Highlights:
      • arXiv:2609.09363v1 Announce Type: new
      • Abstract: Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or d…
      • We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict)
      • To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on loca…

  • Benchmarking Hybrid Deep Research Across Database Querying and Web Search

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

  • However, existing benchmarks evaluate these modalities in isolation, failing to capture the critical “handoff” - the ability to preserve constraints when moving…

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

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

  • The Mutations of Machine Speech

    • Release Time: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09496v1 Announce Type: new.
      • Abstract: Algorithmic outputs now populate the digital environments through which contemporary life is organized.
      • The role of law in facilitating and constituting (rather than merely responding to) these processes is gaining increasing traction across scholarly accounts.
      • This inquiry traces the evolution of algorithmic outputs attending to their legal underpinnings and social implications, surfacing the mutations of machine speech.
    • EN Highlights:
      • arXiv:2609.09496v1 Announce Type: new

  • Abstract: Algorithmic outputs now populate the digital environments through which contemporary life is organized

  • The role of law in facilitating and constituting (rather than merely responding to) these processes is gaining increasing traction across scholarly accounts

  • This inquiry traces the evolution of algorithmic outputs attending to their legal underpinnings and social implications, surfacing the mutations of machine spee…

  • TEFM: Token-Efficient Faithful Modeling for Structured Data

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

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

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

    • Publication Time: 2026-09-10 12:00 Beijing Time
    • Abstract: arXiv:2609.09159v1 Announce Type: new.

  • Abstract: The topological gap $\Delta$ – the excess $H_1$ total persistence of a critical point cloud over a density-matched null – scales as $\Delta \sim L^{d+\eta}$.

  • We derive this analytically: the spectral integral $I(\alpha) = \sum_{k\neq 0} S_{\mathrm{conn}}(k),|k|^{\alpha}$ scales as $L^{2-\alpha-\eta}$ when IR-dominated, giving $I(-2\eta) \sim L^{d+\eta}$.

  • The decomposition $I(-2\eta) = I_0 \cdot I_{\mathrm{shape}}$ separates volume ($I_0 \propto N(1-m^2) \sim L^d$) from anomalous dimension ($I_{\mathrm{shape}} \sim L^{\eta}$); the volume factor accounts for the magnetization-driven per-configuration variance of $\Delta$.

    • EN Key Points:
      • arXiv:2609.09159v1 Announce Type: new
      • Abstract: The topological gap $\Delta$ – the excess $H_1$ total persistence of a critical point cloud over a density-matched null – scales as $\Delta \sim L^{…
      • We derive this analytically: the spectral integral $I(\alpha) = \sum_{k\neq 0} S_{\mathrm{conn}}(k),|k|^{\alpha}$ scales as $L^{2-\alpha-\eta}$ when IR-dominat…
      • The decomposition $I(-2\eta) = I_0 \cdot I_{\mathrm{shape}}$ separates volume ($I_0 \propto N(1-m^2) \sim L^d$) from anomalous dimension ($I_{\mathrm{shape}} \s…

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

    • Publication Time: 2026-09-10 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2609.09162v1 Announce Type: new.
      • Abstract: Physics-informed neural networks (PINNs) provide a mesh-free framework for solving partial differential equations, yet their performance in nonlinear problems is often limited by slow convergence, gradient imbalance, and insufficient resolution to capture localized intermittent structures such as shock waves[1].
      • These issues arise primarily from the use of fixed weights of loss and uniform collocation point distributions, which cannot adapt to the evolving complexity of the solution field during training.

  • To address these challenges, Gradient-Guided Gaussian Adaptive Sampling Physics-Informed Neural Networks (3GAS-PINNs) is proposed in this paper, which combines uniform probability distribution and Gaussian-smoothed probability distribution derived from the spatial gradients of solution, to maintain global constraint satisfaction as well as concentrating collocation points in regions of high gradient.

    • Key Points (EN):
      • arXiv:2609.09162v1 Announce Type: new
      • Abstract: Physics-informed neural networks (PINNs) provide a mesh-free framework for solving partial differential equations, yet their performance in nonlinear…
      • These issues arise primarily from the use of fixed weights of loss and uniform collocation point distributions, which cannot adapt to the evolving complexity of…
      • To address these challenges, Gradient-Guided Gaussian Adaptive Sampling Physics-Informed Neural Networks (3GAS-PINNs) is proposed in this paper, which combines…

  • World-Time Compute with Verified Code World Models

    • Publication Time: 2026-09-10 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.09163v1 Announce Type: new.
      • Abstract: LLMs generalize across a domain only after seeing many real, labeled examples, which most domains lack.
      • We study a way to manufacture it cheaply.
      • When a domain’s dynamics can be written as code, one template instantiates into many world models: executable, verifiable programs over symbolic state, each an inexhaustible source of exactly-labeled trajectories.
    • Key Points (EN):
      • arXiv:2609.09163v1 Announce Type: new
      • Abstract: LLMs generalize across a domain only after seeing many real, labeled examples, which most domains lack
      • We study a way to manufacture it cheaply
      • When a domain’s dynamics can be written as code, one template instantiates into many world models: executable, verifiable programs over symbolic state, each an…

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Summary: [To be translated] - arXiv:2609.09240v1 Announce Type: new.
      • Abstract: Ultra-low-bit language models promise reductions in storage and memory traffic, but a nominal “1.58-bit” label does not specify the deployed representation or its execution cost.
      • We study a scale-up of an aggressive post-training conversion pipeline from Qwen3-4B to Qwen3-8B.
      • The conversion uses KOTMS rotation, E2M-ATQ adaptive ternarisation, and GPTQ-style error compensation in a weight-only A16 configuration.
    • EN Highlights:
      • arXiv:2609.09240v1 Announce Type: new
      • Abstract: Ultra-low-bit language models promise reductions in storage and memory traffic, but a nominal “1.58-bit” label does not specify the deployed represent…
      • We study a scale-up of an aggressive post-training conversion pipeline from Qwen3-4B to Qwen3-8B
      • The conversion uses KOTMS rotation, E2M-ATQ adaptive ternarisation, and GPTQ-style error compensation in a weight-only A16 configuration

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Summary: [To be translated] - arXiv:2609.09241v1 Announce Type: new.
      • Abstract: Mixture-of-Experts (MoE) architectures have emerged as a powerful paradigm for scaling model capacity while preserving efficient inference in large foundation models.
      • However, most MoE models use a fixed top-$k$ expert selection policy, assigning the same expert budget to every token even when fewer experts may be sufficient.
      • Inference-time dynamic top-$k$ routing can reduce computation without retraining, but existing methods often overlook the distributional shift caused by deviating from the training-time routing configuration.
    • EN Highlights:
      • arXiv:2609.09241v1 Announce Type: new

  • Abstract: Mixture-of-Experts (MoE) architectures have emerged as a powerful paradigm for scaling model capacity while preserving efficient inference in large fo…

    • However, most MoE models use a fixed top-$k$ expert selection policy, assigning the same expert budget to every token even when fewer experts may be sufficient
    • Inference-time dynamic top-$k$ routing can reduce computation without retraining, but existing methods often overlook the distributional shift caused by deviati…

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

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

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [Pending Translation] - arXiv:2609.09257v1 Announce Type: new.

  • Abstract: Sensor-based AI systems are rarely operated under the conditions under which they were trained: devices, personnel and recording epochs change, and each change degrades performance in ways a random train-test split cannot reveal.

  • We propose a staged, accountable evaluation protocol that treats the evaluation of a deployed model as a measurement with declared reference levels and a quantified uncertainty.

  • Four cumulative generalisation stages hold out devices, subjects and time.

  • EN Key Points:

    • arXiv:2609.09257v1 Announce Type: new
    • Abstract: Sensor-based AI systems are rarely operated under the conditions under which they were trained: devices, personnel and recording epochs change, and ea…
    • We propose a staged, accountable evaluation protocol that treats the evaluation of a deployed model as a measurement with declared reference levels and a quanti…
    • Four cumulative generalisation stages hold out devices, subjects and time

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

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

  • Abstract: Decision trees are prized for their interpretability and strong performance on tabular data, but popular greedy top-down induction algorithms can yiel…

    • Optimal decision tree methods address this through global optimization, yet remain restricted to axis-aligned threshold splits, which limit the expressivity of…
    • Shape Generalized Trees (SGTs) generalize threshold splits to learnable univariate shape functions, improving expressivity and enabling more compact trees

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

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

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

    • Published: 2026-09-10 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2609.09370v1 Announce Type: new.

  • Abstract: Discrete black-box optimization is often addressed using approaches such as Sequential Model-Based Optimization (SMBO), which aims to improve sample efficiency by fitting surrogate models that approximate a costly objective function over a discrete search space.

  • In many real-world problems, the set of feasible inputs is often given by logical constraints known in advance.

  • However, existing surrogate modeling techniques generally fail to capture the symbolic rules governing feasibility in discrete input spaces.

  • Key Points:

    • arXiv:2609.09370v1 Announce Type: new
    • Abstract: Discrete black-box optimization is often addressed using approaches such as Sequential Model-Based Optimization (SMBO), which aims to improve sample e…
    • In many real-world problems, the set of feasible inputs is often given by logical constraints known in advance
    • However, existing surrogate modeling techniques generally fail to capture the symbolic rules governing feasibility in discrete input spaces