{
  "title": "2026-10-06 AI Daily | OpenAI Discloses EU Text Watermarking Scheme, Agent Evaluation Shifts from 'Getting the Right Answer' to 'Accomplishing the Task'",
  "url": "https://miaok.ong/en/ai-daily/ai-daily-2026-10-06/",
  "date": "2026-10-06T07:00:00+08:00",
  "lastmod": "2026-10-06T07:00:00+08:00",
  "type": "ai-daily",
  "kind": "page",
  "language": "en",
  "description": "OpenAI is simultaneously advancing compliance and commercialization, disclosing its EU text provenance and watermarking boundaries, while also expanding its ChatGPT ad measurement and brand safety tools. The focus of agent evaluation is shifting from the correctness of answers to multi-turn memory, user simulation, and closed-loop execution. On the security front, a high-risk macOS vulnerability is being actively exploited, and discussions are intensifying regarding the dilution of advantages at the model layer.",
  "keywords": null,
  "tags": [],
  "categories": [],
  "author": "Mark (Miao) Kong",
  "image": "https://miaok.ong/images/avatar.jpg",
  "content": "\u003ch1 id=\"2026-10-06-ai-daily--openai-discloses-eu-text-watermarking-plan-agent-evaluation-shifts-from-getting-it-right-to-getting-it-done\"\u003e\n  2026-10-06 AI Daily | OpenAI Discloses EU Text Watermarking Plan, Agent Evaluation Shifts from \u0026ldquo;Getting it Right\u0026rdquo; to \u0026ldquo;Getting it Done\u0026rdquo;\n  \u003ca class=\"heading-link\" href=\"#2026-10-06-ai-daily--openai-discloses-eu-text-watermarking-plan-agent-evaluation-shifts-from-getting-it-right-to-getting-it-done\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h1\u003e\n\u003cblockquote\u003e\n\u003cp\u003eOpenAI is advancing compliance and commercialization in parallel, disclosing its EU plan for text provenance and watermarking boundaries while expanding ad measurement and brand safety tools for ChatGPT. The focus of agent evaluation is shifting from \u0026ldquo;getting it right\u0026rdquo; to multi-turn memory, user simulation, and closed-loop execution. On the security front, a high-risk macOS vulnerability is being actively exploited, and discussions about the diminishing advantage of the model layer are intensifying.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch2 id=\"-in-depth-guide-to-this-issues-watch-list\"\u003e\n  📖 In-depth Guide to This Issue\u0026rsquo;s Watch List\n  \u003ca class=\"heading-link\" href=\"#-in-depth-guide-to-this-issues-watch-list\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003cp\u003eThere are three main themes to watch today. First, OpenAI\u0026rsquo;s parallel push on compliance and commercialization: a blog post disclosed the EU text provenance and watermarking plan, clarifying the scope of watermarking, detection methods, and a researcher-first access mechanism. Simultaneously, ChatGPT introduced new visual ad formats and expanded measurement, attribution, and brand safety tools. Second, agent evaluation is shifting from \u0026ldquo;getting it right\u0026rdquo; to \u0026ldquo;getting it done.\u0026rdquo; XiangqiBench uses a closed-loop Chinese chess task to emphasize execution and verification, APDMem proposes query-adaptive long-term memory, and CUEing User Simulators calibrate multi-turn user simulations. Together, these point to multi-turn interaction, memory, and closed-loop tasks as key for the next phase of evaluation. Third, signals regarding security and competitive moats are noteworthy. An article titled \u0026ldquo;The Billion Dollar AI Advantage Is Disappearing\u0026rdquo; discusses the dilution of advantages at the model layer. Additionally, there are warnings about a high-risk macOS vulnerability being actively exploited, which engineering teams should monitor.\u003c/p\u003e\n\u003ch2 id=\"-quick-takes-on-ai-hot-topics-from-x\"\u003e\n  🌐 Quick Takes on AI Hot Topics from X\n  \u003ca class=\"heading-link\" href=\"#-quick-takes-on-ai-hot-topics-from-x\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003ch3 id=\"topic-1-ai-video-of-jensen-huang-as-michael-jackson-wows-crowds\"\u003e\n  Topic 1: AI Video of Jensen Huang as Michael Jackson Wows Crowds\n  \u003ca class=\"heading-link\" href=\"#topic-1-ai-video-of-jensen-huang-as-michael-jackson-wows-crowds\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCategory: AI · Other\u003c/li\u003e\n\u003cli\u003eSummary: Trending: 13 hours ago, Related posts: 5,200\u003c/li\u003e\n\u003cli\u003eWhat happened: An AI-generated video swapping NVIDIA CEO Jensen Huang with the likeness of Michael Jackson amazed a live audience.\u003c/li\u003e\n\u003cli\u003eWhy it matters: The video showcases new progress in generative AI for realistic human synthesis and entertainment content creation. It also raises concerns about celebrity likeness rights and the risks of deepfakes.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: Discussions on X focused on the video\u0026rsquo;s realism and creative appeal. The point of contention is whether such AI face-swapping should be seen as an innovative form of expression or a potential threat to likeness rights and information authenticity.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-2-openais-dots-ai-agents-launch-with-daily-improvement-pledge\"\u003e\n  Topic 2: OpenAI\u0026rsquo;s Dots AI Agents Launch with Daily Improvement Pledge\n  \u003ca class=\"heading-link\" href=\"#topic-2-openais-dots-ai-agents-launch-with-daily-improvement-pledge\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCategory: AI · News\u003c/li\u003e\n\u003cli\u003eSummary: Trending: 2 days ago, Related posts: 51,000\u003c/li\u003e\n\u003cli\u003eWhat happened: OpenAI launched AI agents named Dots, pledging that the product will be improved daily.\u003c/li\u003e\n\u003cli\u003eWhy it matters: This marks OpenAI\u0026rsquo;s accelerated entry into the AI agent race, pushing high-frequency iteration and agent automation to the forefront of market competition.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: The main topics on X are Dots\u0026rsquo; actual autonomous capabilities, whether the daily improvement pledge can be fulfilled, and how it differs from existing AI agent products.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-3-trump-launches-super-intelligence-force-after-tech-accord\"\u003e\n  Topic 3: Trump Launches Super Intelligence Force After Tech Accord\n  \u003ca class=\"heading-link\" href=\"#topic-3-trump-launches-super-intelligence-force-after-tech-accord\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCategory: AI · News\u003c/li\u003e\n\u003cli\u003eSummary: Trending: 1 day ago, Related posts: 105,000\u003c/li\u003e\n\u003cli\u003eWhat happened: After a tech accord was reached, Trump announced the launch of a \u0026ldquo;Super Intelligence Force\u0026rdquo; to advance the strategic deployment of artificial intelligence.\u003c/li\u003e\n\u003cli\u003eWhy it matters: This move signifies that superintelligence is being integrated into the U.S. national strategy and security framework, potentially accelerating the institutionalization of AI in military, governance, and industrial competition.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: Discussions on X are centered on the specific functions and authority of this force, its relationship with the White House AI accord and the new AI chief, and disagreements over AI safety, ethics, and the centralization of power.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-4-openai-launches-28-day-codex-improvement-sprint\"\u003e\n  Topic 4: OpenAI Launches 28-Day Codex Improvement Sprint\n  \u003ca class=\"heading-link\" href=\"#topic-4-openai-launches-28-day-codex-improvement-sprint\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCategory: AI · News\u003c/li\u003e\n\u003cli\u003eSummary: Trending: 1 day ago, Related posts: 40,000\u003c/li\u003e\n\u003cli\u003eWhat happened: The head of OpenAI Codex announced a 28-day improvement sprint, promising to release one clear improvement for most Codex and ChatGPT Work users every day, or else provide a usage reset to all users.\u003c/li\u003e\n\u003cli\u003eWhy it matters: This is significant for the AI field as it shows a leading AI company using public commitments and penalty mechanisms to address stability and user experience issues in the competitive developer tools market. This could raise the bar for product iteration pace and user expectations for accountability.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: The debate on X includes: supporters who are focused on simplification, efficiency gains, and whether the verifiable daily update promise can be kept. Critics, drawing on the views of a former OpenAI safety report author, worry that a constant sprint culture could squeeze out time for safety checks and reflection, suggesting that speed might be prioritized over quality and safety.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-5-tesla-expands-cybercab-robotaxi-fleet-in-texas-with-safety-first\"\u003e\n  Topic 5: Tesla Expands Cybercab Robotaxi Fleet in Texas with Safety First\n  \u003ca class=\"heading-link\" href=\"#topic-5-tesla-expands-cybercab-robotaxi-fleet-in-texas-with-safety-first\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003eCategory: AI · Entertainment\u003c/li\u003e\n\u003cli\u003eSummary: Trending: 1 day ago, Related posts: 15,000\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eWhat it is\u003c/strong\u003e: Tesla is expanding its Cybercab Robotaxi self-driving taxi fleet in Texas, emphasizing a safety-first approach to advancing the service.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eWhy it matters\u003c/strong\u003e: This marks a significant step for autonomous driving from testing to large-scale commercial operation, setting a benchmark for verifying the safety, cost, and regulatory compliance of AI in real-world traffic scenarios.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eDiscussion summary\u003c/strong\u003e: The main focus on X revolves around the fleet size and safety record. Supporters highlight the zero active collisions and cost-efficiency of the Cybercab. Skeptics, however, point out that most data comes from testers rather than official sources, and discuss the transition pace between Model Y and Cybercab, as well as signals of expansion beyond Texas.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch4 id=\"todays-ai-public-opinion-summary-on-x\"\u003e\n  Today\u0026rsquo;s AI Public Opinion Summary on X\n  \u003ca class=\"heading-link\" href=\"#todays-ai-public-opinion-summary-on-x\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h4\u003e\n\u003cp\u003eToday\u0026rsquo;s main AI discourse shows leading companies and national powers jointly accelerating the large-scale implementation of generative AI, intelligent agents, and autonomous driving, using high-frequency iterations, public commitments, and strategic deployments to seize a competitive edge. The consensus in the discussion is that AI capabilities are rapidly entering real-world scenarios, bringing efficiency and innovative value. However, disagreements are focused on whether \u0026ldquo;speed and safety are out of balance\u0026rdquo;: one side welcomes daily improvements, commercial expansion, and the institutionalization of superintelligence, while the other side worries that a \u0026ldquo;sprint\u0026rdquo; culture squeezes out safety reviews, that deepfakes infringe on portrait rights and information authenticity, and that official verification for autonomous driving and superintelligence is insufficient. The potential risks are that prioritizing competition and efficiency could exacerbate the misuse of deepfakes, lead to an over-concentration of governance power, and blur safety responsibilities in real-world traffic and critical decision-making.\u003c/p\u003e\n\u003ch2 id=\"-influencer-insights\"\u003e\n  💡 Influencer Insights\n  \u003ca class=\"heading-link\" href=\"#-influencer-insights\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003cblockquote\u003e\n\u003cp\u003eNo influencer insights today. In-depth content from the Watch List is recommended.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch2 id=\"-appendix-todays-watch-list-update-source-list\"\u003e\n  📚 Appendix: Today\u0026rsquo;s Watch List Update Source List\n  \u003ca class=\"heading-link\" href=\"#-appendix-todays-watch-list-update-source-list\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h2\u003e\n\u003cblockquote\u003e\n\u003cp\u003eTimeframe: Last 3 days; covers 22 sources; 34 updates in total\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003ch3 id=\"stratechery-by-ben-thompson-a_full\"\u003e\n  Stratechery by Ben Thompson (A_full)\n  \u003ca class=\"heading-link\" href=\"#stratechery-by-ben-thompson-a_full\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003e\u003ca href=\"https://stratechery.com/2026/apple-and-a-hackers-future/\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eApple and a Hacker’s Future\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 18:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: [TO BE TRANSLATED] - \u003cstrong\u003eListen to this \u003cstrong\u003epost\u003c/strong\u003e:\u003c/strong\u003e.\n\u003cul\u003e\n\u003cli\u003e\n\u003cblockquote\u003e\n\u003cp\u003eDutch officials have warned that a high-severity macOS vulnerability that allows attackers to execute malicious code is under active exploitation.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003c/li\u003e\n\u003cli\u003e“The NCSC has received a notification indicating that active abuse of this vulnerability has been observed on multiple systems on which port 5900 was accessible from the Internet,” the Netherlands National Cyber Security Centrum warned earlier this week.\u003c/li\u003e\n\u003cli\u003e“In all these cases, root had been accessed on the affected system and a Monero crypto miner had been placed.”.\u003c/li\u003e\n\u003cli\u003e\n\u003cblockquote\u003e\n\u003cp\u003eThe vulnerability, tracked as CVE-2026-65400, received a patch from Apple last week for macOS Tahoe, Sequoia, and Sonoma.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003eListen to this post :\u003c/li\u003e\n\u003cli\u003eLog in to listen\u003c/li\u003e\n\u003cli\u003eMy computer got hacked, which is always embarrassing to admit, because it was my fault; the vulnerability that was exploited is detailed in this Ars Technica st…\u003c/li\u003e\n\u003cli\u003eDutch officials have warned that a high-severity macOS vulnerability that allows attackers to execute malicious code is under active exploitation\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"openai-blog-a_full\"\u003e\n  OpenAI Blog (A_full)\n  \u003ca class=\"heading-link\" href=\"#openai-blog-a_full\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://openai.com/index/eu-text-provenance\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eOur approach to EU text provenance rules\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 23:00 Beijing Time\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eSummary: How OpenAI is approaching text watermarking under EU rules.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eLearn where watermarks apply, how detection works, and why access starts with researchers.\u003c/li\u003e\n\u003cli\u003eHow OpenAI is approaching text watermarking under EU rules Learn where watermarks apply, how detection works, and why access starts with researchers.\u003c/li\u003e\n\u003cli\u003eKey Points (EN):\n\u003cul\u003e\n\u003cli\u003eHow OpenAI is approaching text watermarking under EU rules\u003c/li\u003e\n\u003cli\u003eLearn where watermarks apply, how detection works, and why access starts with researchers.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://openai.com/index/new-chatgpt-ads-format-and-measurement\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eBuilding advertising for the way people use AI\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 18:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: OpenAI introduces a new visual ad format in ChatGPT and expands measurement tools, attribution partnerships, and brand suitability for advertisers.\n\u003cul\u003e\n\u003cli\u003eThis piece from OpenAI Blog explains how Building advertising for the way people use AI shapes the broader AI and infrastructure landscape.\u003c/li\u003e\n\u003cli\u003eIt also surfaces practical implications for founders, operators, and investors following Building advertising for the way people use AI.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eKey Points (EN):\n\u003cul\u003e\n\u003cli\u003eOpenAI introduces a new visual ad format in ChatGPT and expands measurement tools, attribution partnerships, and brand suitability for advertisers.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"two-minute-papers-b_introsearch\"\u003e\n  Two Minute Papers (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#two-minute-papers-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003e\u003ca href=\"https://www.youtube.com/watch?v=ZHVNTTKu9fU\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eThe Billion Dollar AI Advantage Is Disappearing\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 17:53 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: ❤️ Check out Lambda here and sign up for their GPU Cloud:.\n\u003cul\u003e\n\u003cli\u003eAdam Bridges, B Shang, Carlos Galarza, Christian Ahlin, Eric Tyson, Juan Benet, Lukas Biewald, Michael Tedder, Owen Skarpness, Ryan Stankye, Shawn Becker, Steef, Taras Bobrovytsky, Tazaur Sagenclaw, Tybie Fitzhugh, Ueli Gallizzi.\u003c/li\u003e\n\u003cli\u003eThe Billion Dollar AI Advantage Is Disappearing.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eKey Points (EN):\n\u003cul\u003e\n\u003cli\u003e❤️ Check out Lambda here and sign up for their GPU Cloud:\u003c/li\u003e\n\u003cli\u003e📝 Sonnet 5.5:\u003c/li\u003e\n\u003cli\u003e🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eAdam Bridges, B Shang, Carlos Galarza, Christian Ahlin, Eric Tyson, Juan Benet, Lukas Biewald, Michael Tedder, Owen Skarpness, Ryan Stankye, Shawn Becker, Steef…\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"arxiv-csai-b_introsearch\"\u003e\n  ArXiv cs.AI (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#arxiv-csai-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02260\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eMintFlow: Minimal Trajectory Intervention for Constrained Flow Matching\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: [Translation Pending] - arXiv:2610.02260v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Flow matching models excel at generative modeling, and many downstream applications require their samples to satisfy prescribed constraints, such as observed measurements and physical laws.\u003c/li\u003e\n\u003cli\u003eHowever, existing constrained samplers often face a trade-off: \\textit{enforcing constraints can substantially displace samples from the pretrained data distribution}.\u003c/li\u003e\n\u003cli\u003eTo address this trade-off, we introduce \\textbf{MintFlow}, a training-free constrained sampling framework that formulates constraint enforcement as a minimal intervention on the pretrained flow trajectory.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eKey Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02260v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Flow matching models excel at generative modeling, and many downstream applications require their samples to satisfy prescribed constraints, such as o…\u003c/li\u003e\n\u003cli\u003eHowever, existing constrained samplers often face a trade-off: \\textit{enforcing constraints can substantially displace samples from the pretrained data distrib…\u003c/li\u003e\n\u003cli\u003eTo address this trade-off, we introduce \\textbf{MintFlow}, a training-free constrained sampling framework that formulates constraint enforcement as a minimal in…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02267\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eFast Models, Slow Evidence: A Paired and Self-Audited Evaluation of System-1 Decision Models for LLM Agent Harnesses\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: [Translation Pending] - arXiv:2610.02267v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Agent harnesses make many small, typed decisions per task: which model to call, which tool to use, whether retrieved text is relevant, whether an input carries an injection.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eSystem-1 decision models answer such questions in a single forward pass with class probabilities, promising large cost and latency savings over LLM calls.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe present a paired evaluation of an open-weight (Laya) and a hosted (Jev) System-1 model on 11 agent decision points built from 18 public sources: 7,283 base cases plus 6,640 robustness variants, with byte-identical inputs, paired tests, and cross-hardware and cross-day reproducibility checks.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02267v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Agent harnesses make many small, typed decisions per task: which model to call, which tool to use, whether retrieved text is relevant, whether an inpu…\u003c/li\u003e\n\u003cli\u003eSystem-1 decision models answer such questions in a single forward pass with class probabilities, promising large cost and latency savings over LLM calls\u003c/li\u003e\n\u003cli\u003eWe present a paired evaluation of an open-weight (Laya) and a hosted (Jev) System-1 model on 11 agent decision points built from 18 public sources: 7,283 base c…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02281\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eThe AI Risk Observatory: What Can We Learn from AI Disclosures in Annual Reports About Societal Resilience?\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [To be translated] - arXiv:2610.02281v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Societal resilience research relies on access to useful and actionable data, which motivates our main research question: Can annual reports, processed at scale with LLMs, provide a useful signal about how companies disclose their response to AI?\u003c/li\u003e\n\u003cli\u003eWe test this by applying a reproducible two-stage classification pipeline to 9,821 annual reports from 1,362 UK listed companies (2020-2025, with partial 2026 data).\u003c/li\u003e\n\u003cli\u003eWe first validate the method against 474 human-annotated passages, finding high recall and moderate label-level agreement.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02281v1 Announce Type: new\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02300\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eKeep It CALM: Analyzing the Limits of Global Unsafety in Text-to-Image Generation\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [To be translated] - arXiv:2610.02300v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Training-free safeguards for text-to-image generation often rely on a reusable safety signal, such as an unsafe direction or global toxic subspace, applied broadly across prompts.\u003c/li\u003e\n\u003cli\u003eWe provide a controlled geometric analysis of this global-unsafety assumption and reveal a consistent coverage-selectivity trade-off: compact unsafe subspaces fail to cover heterogeneous unsafe semantics, whereas broader aggregation increasingly distorts safety-adjacent benign prompts.\u003c/li\u003e\n\u003cli\u003eMotivated by this finding, we propose CALM (Counterfactual Adaptive Local Modulation), a training-free safeguard that replaces uniform global removal with prompt-local counterfactual correction.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02300v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Training-free safeguards for text-to-image generation often rely on a reusable safety signal, such as an unsafe direction or global toxic subspace, ap…\u003c/li\u003e\n\u003cli\u003eWe provide a controlled geometric analysis of this global-unsafety assumption and reveal a consistent coverage-selectivity trade-off: compact unsafe subspaces f…\u003c/li\u003e\n\u003cli\u003eMotivated by this finding, we propose CALM (Counterfactual Adaptive Local Modulation), a training-free safeguard that replaces uniform global removal with promp…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02330\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eChoosing Before Acting: Comparative Value Estimation for Long-Horizon Tool-Use Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: [TO BE TRANSLATED] - arXiv:2610.02330v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Large language models (LLMs) rely on long-horizon tool invocation sequences for complex tasks, where each invocation can alter the task state and condition subsequent decisions.\u003c/li\u003e\n\u003cli\u003eIn long-horizon tool use, final-outcome rewards provide weak credit assignment over long interaction traces.\u003c/li\u003e\n\u003cli\u003eStep-level rewards can offer more targeted feedback, but obtaining reliable step supervision often requires human or LLM judgment, or additional rollouts to estimate the downstream effect of an intermediate decision.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02330v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Large language models (LLMs) rely on long-horizon tool invocation sequences for complex tasks, where each invocation can alter the task state and cond…\u003c/li\u003e\n\u003cli\u003eIn long-horizon tool use, final-outcome rewards provide weak credit assignment over long interaction traces\u003c/li\u003e\n\u003cli\u003eStep-level rewards can offer more targeted feedback, but obtaining reliable step supervision often requires human or LLM judgment, or additional rollouts to est…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02331\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eWorld Editing: Intervening on Executable Worlds at Increasing Depth\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: [TO BE TRANSLATED] - arXiv:2610.02331v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an existing executable world remains underexplored.\u003c/li\u003e\n\u003cli\u003eWe formulate world editing as intervening on an existing world while preserving properties that should remain unchanged, and introduce intervention depth as an axis describing how strongly an edit couples world entities, dynamics, and systems.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe instantiate this capability through industry-grade game modding and introduce IGMWorld, together with IGMBench, a benchmark of 110 tasks and over 1.1K executable state and behavioral criteria across Minecraft and Terraria.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eKey Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02331v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Interactive world models are increasingly capable of generating environments and acting within them, yet deliberately editing an existing executable w…\u003c/li\u003e\n\u003cli\u003eWe formulate world editing as intervening on an existing world while preserving properties that should remain unchanged, and introduce intervention depth as an…\u003c/li\u003e\n\u003cli\u003eWe instantiate this capability through industry-grade game modding and introduce IGMWorld, together with IGMBench, a benchmark of 110 tasks and over 1.1K execut…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02342\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eA Multi Method Importance and Performance Efficiency Analysis of Topological Metrics for Natural Visibility Graph Based Cyber Attack Detection\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: arXiv:2610.02342v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Natural Visibility Graph (NVG) based analysis characterizes network traffic through topological descriptors reflecting different structural properties.\u003c/li\u003e\n\u003cli\u003eHowever, not all descriptors contribute equally to cyber-attack classification, and extracting a large metric set can increase computational cost.\u003c/li\u003e\n\u003cli\u003eThis study evaluates 21 NVG derived topological metrics and investigates whether a compact subset can preserve classification capability while improving computational efficiency.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eKey Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02342v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Natural Visibility Graph (NVG) based analysis characterizes network traffic through topological descriptors reflecting different structural properties\u003c/li\u003e\n\u003cli\u003eHowever, not all descriptors contribute equally to cyber-attack classification, and extracting a large metric set can increase computational cost\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eThis study evaluates 21 NVG derived topological metrics and investigates whether a compact subset can preserve classification capability while improving computa…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02351\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eDeReAct: Decomposed Reasoning and Acting for Reliable AI Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Pending Translation] - arXiv:2610.02351v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: ReAct-based agents typically rely on a single LLM policy to propose actions, interact with the environment, and decide when a task is complete.\u003c/li\u003e\n\u003cli\u003eThis coupling makes action authorization and completion control difficult to enforce independently, allowing errors to propagate and unsupported completion claims to terminate execution.\u003c/li\u003e\n\u003cli\u003eWe introduce DeReAct, a modular agent architecture that externalizes two gating policies: a Critic that validates proposed actions before execution, and a Context Manager that reconstructs an environment-supported \\textsc{State} and certifies task completion.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02351v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: ReAct-based agents typically rely on a single LLM policy to propose actions, interact with the environment, and decide when a task is complete\u003c/li\u003e\n\u003cli\u003eThis coupling makes action authorization and completion control difficult to enforce independently, allowing errors to propagate and unsupported completion clai…\u003c/li\u003e\n\u003cli\u003eWe introduce DeReAct, a modular agent architecture that externalizes two gating policies: a Critic that validates proposed actions before execution, and a Conte…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02372\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eTraversing the Satisfaction-Diversity Frontier in Text-to-Image Diffusion\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Pending Translation] - arXiv:2610.02372v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Text-to-image generation enables users to explore several images generated from the same prompt.\u003c/li\u003e\n\u003cli\u003eFor these generated images to be useful, each one must reflect the user\u0026rsquo;s preferences, measured by a learned reward, and differ visually from the others to maintain diversity.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eExisting methods are limited: they either address reward and diversity separately or combine them in one aggregate score, enabling high diversity to offset low rewards.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02372v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Text-to-image generation enables users to explore several images generated from the same prompt\u003c/li\u003e\n\u003cli\u003eFor these generated images to be useful, each one must reflect the user\u0026rsquo;s preferences, measured by a learned reward, and differ visually from the others to main…\u003c/li\u003e\n\u003cli\u003eExisting methods are limited: they either address reward and diversity separately or combine them in one aggregate score, enabling high diversity to offset low…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02378\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eTHPL: A Vision-to-Language Decision Support Framework for Rainbow Trout Feeding Management in RAS\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Translation Pending] - arXiv:2610.02378v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: In Recirculating Aquaculture Systems (RAS), precision feeding is critical for minimizing costs and improving fish welfare.\u003c/li\u003e\n\u003cli\u003eHowever, existing methods lack cognitive alignment between fish behaviors and management knowledge, impeding translation into executable, interpretable feeding decisions.\u003c/li\u003e\n\u003cli\u003eTo address this, we propose THPL, a generative feeding decision framework tailored for rainbow trout (Oncorhynchus mykiss) in RAS.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02378v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: In Recirculating Aquaculture Systems (RAS), precision feeding is critical for minimizing costs and improving fish welfare\u003c/li\u003e\n\u003cli\u003eHowever, existing methods lack cognitive alignment between fish behaviors and management knowledge, impeding translation into executable, interpretable feeding…\u003c/li\u003e\n\u003cli\u003eTo address this, we propose THPL, a generative feeding decision framework tailored for rainbow trout (Oncorhynchus mykiss) in RAS\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"arxiv-cscl-b_introsearch\"\u003e\n  ArXiv cs.CL (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#arxiv-cscl-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02293\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eHakemBench: A Turkish Benchmark of Typed Decisions\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAbstract: arXiv:2610.02293v1 Announce Type: new.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eAbstract: HakemBench is a Turkish benchmark of typed decisions, in which the model under test reads a text, a question and a fixed set of options and returns a probability for every option.\u003c/li\u003e\n\u003cli\u003eVersion 1.0 is released fully open under CC BY 4.0, with 2,346 items and 4,275 choice, yes/no and score questions in seven tracks (fact-check triage, education, guardrails, legal routing, moderation, spam and phishing, and customer support).\u003c/li\u003e\n\u003cli\u003eOne harness scores decision quality (macro F1), calibration (from the normalised Brier score) and selective automation (from the normalised area under the generalised risk-coverage curve), combines them by a geometric mean and reports intervals from 2,000 bootstrap draws; probes for option order, paraphrase, English translation and substituted names are reported alongside.\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02293v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: HakemBench is a Turkish benchmark of typed decisions, in which the model under test reads a text, a question and a fixed set of options and returns a…\u003c/li\u003e\n\u003cli\u003eVersion 1.0 is released fully open under CC BY 4.0, with 2,346 items and 4,275 choice, yes/no and score questions in seven tracks (fact-check triage, education,…\u003c/li\u003e\n\u003cli\u003eOne harness scores decision quality (macro F1), calibration (from the normalised Brier score) and selective automation (from the normalised area under the gener…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02425\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eFinding the Move Is Not Winning the Game: XiangqiBench for Closed-Loop Evaluation of LLM Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: arXiv:2610.02425v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Static evaluations credit a language model for naming the right move, but an agent must carry a plan through to a verified outcome while an opponent responds.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe introduce XiangqiBench, an executable benchmark that measures this difference in Chinese chess: starting from 119 tactical endgames with forced mates supported by engine or checks-only search, an LLM agent must deliver checkmate against an engine defender.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAn interactive REPL interface separates real moves, state queries, and forward simulation, and we record 8,568 multi-turn trajectories from 12 frontier LLMs under two observation protocols.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02425v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Static evaluations credit a language model for naming the right move, but an agent must carry a plan through to a verified outcome while an opponent r…\u003c/li\u003e\n\u003cli\u003eWe introduce XiangqiBench, an executable benchmark that measures this difference in Chinese chess: starting from 119 tactical endgames with forced mates support…\u003c/li\u003e\n\u003cli\u003eAn interactive REPL interface separates real moves, state queries, and forward simulation, and we record 8,568 multi-turn trajectories from 12 frontier LLMs und…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02444\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eCounterexample Generation via Per-Theorem Symbolic Verifiers: When Imitation Hurts and Reinforcement Repairs\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Translation Pending] - arXiv:2610.02444v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Large language models often solve a theorem forward yet fail to disprove a closely related false one: a falsification gap that supervised fine-tuning does not close and can actively worsen.\u003c/li\u003e\n\u003cli\u003eWe frame counterexample generation as constrained witness emission against a deterministic per-theorem Python verifier, and release SymCE, a corpus of 4,707 false undergraduate-algebra and real-analysis conjectures, each paired with executable verifiers.\u003c/li\u003e\n\u003cli\u003eThe verifier also serves as the reward function, making SymCE a training environment.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02444v1 Announce Type: new\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAbstract: Large language models often solve a theorem forward yet fail to disprove a closely related false one: a falsification gap that supervised fine-tuning…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe frame counterexample generation as constrained witness emission against a deterministic per-theorem Python verifier, and release SymCE, a corpus of 4,707 fal…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eThe verifier also serves as the reward function, making SymCE a training environment\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02455\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eFinDialogLens: Event Extraction over Multi-Party Dialogue for Missed-Trade Identification in Financial Chatrooms\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Pending Translation] - arXiv:2610.02455v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Multi-party financial chatrooms are vital for sales-and-trading professionals, but their complexity makes manual recovery of missed trades infeasible: each Request for Quote (RFQ) is an event whose final price and trade outcome appear many messages after the RFQ-trigger message (the inquiry message), interleaved with concurrent RFQs from other participants.\u003c/li\u003e\n\u003cli\u003eWe cast this as event extraction (EE) over multi-party dialogue and present FinDialogLens, a hybrid LLM pipeline in which compact fine-tuned classifiers act as inference-time scaffolds: they detect RFQ-triggers and price/trade outcome metadata, an RFQ-Level Module segments per-event RFQ windows, and a Trade Engine fills argument roles.\u003c/li\u003e\n\u003cli\u003eWith GPT-4o, FinDialogLens reaches 92.1% and 94.3% accuracy on final price and trade outcome, respectively, outperforming full-chatroom CoT prompting methods; fine-tuned open-source LLMs with as few as 3B parameters achieve comparable performance with modest in-domain data.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02455v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Multi-party financial chatrooms are vital for sales-and-trading professionals, but their complexity makes manual recovery of missed trades infeasible:…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe cast this as event extraction (EE) over multi-party dialogue and present FinDialogLens, a hybrid LLM pipeline in which compact fine-tuned classifiers act as…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWith GPT-4o, FinDialogLens reaches 92.1% and 94.3% accuracy on final price and trade outcome, respectively, outperforming full-chatroom CoT prompting methods; f…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02460\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eCUEing User Simulators: Calibrated User Embeddings for Multi-Turn Benchmarking\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Translation Pending] - arXiv:2610.02460v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Recent benchmarks rely on user simulators to evaluate AI agents in multi-turn interaction.\u003c/li\u003e\n\u003cli\u003eWhile existing simulation techniques demonstrate surface fidelity to human style and behavior, ecologically valid interactive benchmarking also requires alignment in when and how agents fail across simulated and real user populations.\u003c/li\u003e\n\u003cli\u003eWe find that existing simulators lack outcome calibration: agreement with observed success rates and failure patterns when real users interact with the same agent.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02460v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Recent benchmarks rely on user simulators to evaluate AI agents in multi-turn interaction\u003c/li\u003e\n\u003cli\u003eWhile existing simulation techniques demonstrate surface fidelity to human style and behavior, ecologically valid interactive benchmarking also requires alignme…\u003c/li\u003e\n\u003cli\u003eWe find that existing simulators lack outcome calibration: agreement with observed success rates and failure patterns when real users interact with the same age…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02472\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eAPDMem: Agent-Controlled Progressive Disclosure for Query-Adaptive Long-Term Memory\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Translation Pending] - arXiv:2610.02472v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Personalized LLM assistants must recover sparse evidence from long conversation histories across queries of varying complexity.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe introduce APDMem (Agent-controlled Progressive Disclosure Memory), a hierarchical long-term memory architecture that applies progressive disclosure to memory retrieval.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eRather than relying on a flat memory store or fixed retrieval granularity, APDMem represents conversation history as four progressively detailed layers: thematic summaries, personalized key facts, turn-level evidence notes, and raw messages.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eEN Highlights:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02472v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Personalized LLM assistants must recover sparse evidence from long conversation histories across queries of varying complexity\u003c/li\u003e\n\u003cli\u003eWe introduce APDMem (Agent-controlled Progressive Disclosure Memory), a hierarchical long-term memory architecture that applies progressive disclosure to memory…\u003c/li\u003e\n\u003cli\u003eRather than relying on a flat memory store or fixed retrieval granularity, APDMem represents conversation history as four progressively detailed layers: themati…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02486\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eFrom Retrieval to Typed Decisions: Calibrated System One Models from Biomedical Sentence Encoders\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Translation pending] - arXiv:2610.02486v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Typed decision models answer schema-constrained questions about a text in one forward pass and return probabilities meant to be thresholded.\u003c/li\u003e\n\u003cli\u003eWe ask whether biomedical sentence encoders trained for retrieval are good starting points for such models.\u003c/li\u003e\n\u003cli\u003eWe present SBERT2S1, which converts Sentence-Transformers encoders into bi-encoder, cross-head (C) and prior-fused residual (PFR) decision models, together with BIODECIDE, a biomedical typed-decision suite, and MEDLINE-S1, 243k training decisions derived from NLM indexing.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02486v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Typed decision models answer schema-constrained questions about a text in one forward pass and return probabilities meant to be thresholded\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe ask whether biomedical sentence encoders trained for retrieval are good starting points for such models\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe present SBERT2S1, which converts Sentence-Transformers encoders into bi-encoder, cross-head (C) and prior-fused residual (PFR) decision models, together with…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02529\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eA generative-informed neuro-symbolic framework for syntactic ambiguity resolution: Evidence from Arabic DPs\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Awaiting Translation] - arXiv:2610.02529v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Syntactic ambiguity poses a persistent challenge for Arabic NLP, particularly in morphologically rich nominal constructions where multiple structu6ral interpretations may be compatible with the same surface sequence.\u003c/li\u003e\n\u003cli\u003eThis study proposes a generatively informed neuro-symbolic framework for resolving structural ambiguity in Modern Standard Arabic (MSA) DPs.\u003c/li\u003e\n\u003cli\u003eThe framework integrates generative syntactic notions with AraBERT by representing ambiguity as a candidate-based decision task in which linguistically motivated alternatives are explicitly constructed and evaluated through candidate-conditioned input representations.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02529v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Syntactic ambiguity poses a persistent challenge for Arabic NLP, particularly in morphologically rich nominal constructions where multiple structu6ral…\u003c/li\u003e\n\u003cli\u003eThis study proposes a generatively informed neuro-symbolic framework for resolving structural ambiguity in Modern Standard Arabic (MSA) DPs\u003c/li\u003e\n\u003cli\u003eThe framework integrates generative syntactic notions with AraBERT by representing ambiguity as a candidate-based decision task in which linguistically motivate…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02549\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eEvaluating Multi-Dimensional Generalization of Large Language Models in Temporal Extraction Tasks\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Awaiting Translation] - arXiv:2610.02549v1 Announce Type: new.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAbstract: Time and event expression extraction are fundamental temporal reasoning tasks, but the problem remains difficult due to annotation ambiguity, domain sensitivity, and unstable model behavior.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eExisting evaluations focus on in-domain performance, offering limited insight into reliability under distribution shifts.\u003c/li\u003e\n\u003cli\u003eWe evaluate multiple model configurations across families, architectures, and reasoning strategies over four dimensions of generalization, examining transfer from base performance, cross-dimensional correlations, and the effects of scale, architecture, and prompting.\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02549v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Time and event expression extraction are fundamental temporal reasoning tasks, but the problem remains difficult due to annotation ambiguity, domain s…\u003c/li\u003e\n\u003cli\u003eExisting evaluations focus on in-domain performance, offering limited insight into reliability under distribution shifts\u003c/li\u003e\n\u003cli\u003eWe evaluate multiple model configurations across families, architectures, and reasoning strategies over four dimensions of generalization, examining transfer fr…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02612\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eLearning When to Commit from Partial Speech for End-to-End Simultaneous Speech Translation\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: arXiv:2610.02612v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Simultaneous speech translation must emit useful target text before the source is complete while preserving every committed token.\u003c/li\u003e\n\u003cli\u003eWe adapt a full-utterance speech language model using prefix supervision derived from its own complete- and partial-waveform translations, requiring neither transcripts nor human translations.\u003c/li\u003e\n\u003cli\u003eWe compare single-turn forced-prefix and multi-turn append-only decoding, use a confidence threshold to control the inference-time quality\u0026ndash;latency trade-off, and vary the density of training prefixes with a separate synthesis margin.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02612v1 Announce Type: new\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAbstract: Simultaneous speech translation must emit useful target text before the source is complete while preserving every committed token\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe adapt a full-utterance speech language model using prefix supervision derived from its own complete- and partial-waveform translations, requiring neither tra…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe compare single-turn forced-prefix and multi-turn append-only decoding, use a confidence threshold to control the inference-time quality\u0026ndash;latency trade-off, a…\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"arxiv-cslg-b_introsearch\"\u003e\n  ArXiv cs.LG (B_intro+search)\n  \u003ca class=\"heading-link\" href=\"#arxiv-cslg-b_introsearch\"\u003e\n    \u003ci class=\"fa-solid fa-link\" aria-hidden=\"true\" title=\"Link to heading\"\u003e\u003c/i\u003e\n    \u003cspan class=\"sr-only\"\u003eLink to heading\u003c/span\u003e\n  \u003c/a\u003e\n\u003c/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02224\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eHybrid Machine Learning-Assisted Raman Spectroscopy with Generative Feature Augmentation for Pharmaceutical Identification\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eSummary: [Translation pending] - arXiv:2610.02224v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Rapid and reliable identification of pharmaceutical residues is important for safeguarding public health, ensuring food safety, and enabling practical Raman-based screening.\u003c/li\u003e\n\u003cli\u003eIn this study, we propose HyMLRaman, a hybrid Raman spectroscopy framework that combines deep spectral feature extraction, generative models, and classical machine-learning classifiers to identify six pharmaceutical compounds, including amoxicillin, chloramphenicol, ciprofloxacin, tetracycline, ibuprofen, and paracetamol.\u003c/li\u003e\n\u003cli\u003eRaman spectra are converted into spectral images and encoded with several deep neural-network backbones, among which EfficientNet-B3 yields the most effective representation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02224v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Rapid and reliable identification of pharmaceutical residues is important for safeguarding public health, ensuring food safety, and enabling practical…\u003c/li\u003e\n\u003cli\u003eIn this study, we propose HyMLRaman, a hybrid Raman spectroscopy framework that combines deep spectral feature extraction, generative models, and classical mach…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eRaman spectra are converted into spectral images and encoded with several deep neural-network backbones, among which EfficientNet-B3 yields the most effective r…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02225\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eThe Price of Greenwashing: Algorithmic Verification and Market Discipline using Conformal Machine Learning\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [To be translated] - arXiv:2610.02225v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: While corporate sustainability mandates are expanding, the systemic reliance on self-reported emissions data exposes financial markets to pervasive greenwashing.\u003c/li\u003e\n\u003cli\u003eCurrent literature relies heavily on subjective ESG ratings or textual sentiment analysis, leaving a critical econometric gap in objectively quantifying physical climate realities.\u003c/li\u003e\n\u003cli\u003eTo resolve this information asymmetry, we fuse U.S.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02225v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: While corporate sustainability mandates are expanding, the systemic reliance on self-reported emissions data exposes financial markets to pervasive gr…\u003c/li\u003e\n\u003cli\u003eCurrent literature relies heavily on subjective ESG ratings or textual sentiment analysis, leaving a critical econometric gap in objectively quantifying physica…\u003c/li\u003e\n\u003cli\u003eTo resolve this information asymmetry, we fuse U.S\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02248\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eState-Space Unlearning for Non-Stationary Bias in Land Surface Forecasting\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [To be translated] - arXiv:2610.02248v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Operational land surface forecasting systems built on Mamba-family Structured State Space Models absorb non-stationary confounding events (unrecorded irrigation booms, dam-operation shifts, sensor recalibrations) into their state-transition matrices, silently biasing NDVI, LST, and crop phenology predictions long after the physical cause ends.\u003c/li\u003e\n\u003cli\u003eThis paper introduces SSU-LSF (State-Space Unlearning for Land Surface Forecasting), the first machine-unlearning framework purpose-built for geoscientific Mamba-based SSMs.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWe develop EKFac influence functions specialized to the Mamba state matrices via a closed-form matrix-exponential gradient, use spectral-radius-weighted elbow thresholding to localize a temporal confounding footprint $\\Phi$, and apply Hessian-free projected gradient ascent within a KL-divergence trust region augmented by spatial total-variation (TV) regularization.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02248v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Operational land surface forecasting systems built on Mamba-family Structured State Space Models absorb non-stationary confounding events (unrecorded…\u003c/li\u003e\n\u003cli\u003eThis paper introduces SSU-LSF (State-Space Unlearning for Land Surface Forecasting), the first machine-unlearning framework purpose-built for geoscientific Mamb…\u003c/li\u003e\n\u003cli\u003eWe develop EKFac influence functions specialized to the Mamba state matrices via a closed-form matrix-exponential gradient, use spectral-radius-weighted elbow t…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02249\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eNearest-neighbour baselines for fingerprint prediction from MS/MS spectra under different assumptions\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [PENDING TRANSLATION] - arXiv:2610.02249v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: It has recently been shown that nearest-neighbour retrieval provides a strong baseline for molecular fingerprint prediction from MS/MS spectra, with several variants matching or outperforming current deep learning models (Khoo and Barzilay, 2026; Liu et al., 2026; Gupta et al., 2026).\u003c/li\u003e\n\u003cli\u003eImportantly, \u0026ldquo;nearest neighbour\u0026rdquo; encompasses a family of retrieval methods that differ in the information assumed to be available at inference.\u003c/li\u003e\n\u003cli\u003eIn this report, we systematically compare several nearest-neighbour variants and show how these differing assumptions affect performance.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02249v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: It has recently been shown that nearest-neighbour retrieval provides a strong baseline for molecular fingerprint prediction from MS/MS spectra, with s…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eImportantly, \u0026ldquo;nearest neighbour\u0026rdquo; encompasses a family of retrieval methods that differ in the information assumed to be available at inference\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eIn this report, we systematically compare several nearest-neighbour variants and show how these differing assumptions affect performance\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02251\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eRank-Aware Speculative Sampling for Diffusion Draft Trees\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: arXiv:2610.02251v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Speculative sampling accelerates diffusion generation by verifying inexpensive draft states in parallel while preserving the target law.\u003c/li\u003e\n\u003cli\u003eRecent tree-based methods allocate the parallel compute budget more effectively than single-chain drafts, as demonstrated by Diffusion Greedy Rejection Sampling (D-GRS).\u003c/li\u003e\n\u003cli\u003eD-GRS generates $K$ conditionally independent candidates per node, and sequentially tests them in their generation order.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eKey English Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02251v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Speculative sampling accelerates diffusion generation by verifying inexpensive draft states in parallel while preserving the target law\u003c/li\u003e\n\u003cli\u003eRecent tree-based methods allocate the parallel compute budget more effectively than single-chain drafts, as demonstrated by Diffusion Greedy Rejection Sampling…\u003c/li\u003e\n\u003cli\u003eD-GRS generates $K$ conditionally independent candidates per node, and sequentially tests them in their generation order\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02252\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eCounterfactual Predictions in Scientific Emulators Without Controlled Experiments\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: arXiv:2610.02252v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Many scientific questions require reasoning about what was never observed: What if the conditions, interventions, or history had been different?\u003c/li\u003e\n\u003cli\u003eModels can predict accurately on observed data yet fail on such what-if queries when correlated inputs are varied independently.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eA common remedy is to add controlled simulation data in which these factors are explicitly disentangled, but this requires access to a simulator, can be computationally expensive, and inherits the simulator\u0026rsquo;s modeling assumptions.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02252v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Many scientific questions require reasoning about what was never observed: What if the conditions, interventions, or history had been different\u003c/li\u003e\n\u003cli\u003eModels can predict accurately on observed data yet fail on such what-if queries when correlated inputs are varied independently\u003c/li\u003e\n\u003cli\u003eA common remedy is to add controlled simulation data in which these factors are explicitly disentangled, but this requires access to a simulator, can be computa…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02253\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eApproximation Property of Dropout Neural Networks: Sobolev Rates and Confidence Bounds\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [To be translated] - arXiv:2610.02253v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: The universal approximation property of dropout neural networks does not by itself describe the network size required for an accurate random realization.\u003c/li\u003e\n\u003cli\u003eIn this work, we study approximation of the unit ball of $W^{n,\\infty}([0,1]^d)$ by ReLU networks whose edges are retained independently with probability $p$.\u003c/li\u003e\n\u003cli\u003eThe approximation error is measured uniformly over the input domain, and the guarantee holds with probability at least $1-\\delta$ for a single sampled network.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02253v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: The universal approximation property of dropout neural networks does not by itself describe the network size required for an accurate random realizati…\u003c/li\u003e\n\u003cli\u003eIn this work, we study approximation of the unit ball of $W^{n,\\infty}([0,1]^d)$ by ReLU networks whose edges are retained independently with probability $p$\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eThe approximation error is measured uniformly over the input domain, and the guarantee holds with probability at least $1-\\delta$ for a single sampled network\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02254\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eOvercoming Challenges of Interpretive Structural Modeling with Large Language Models\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Pending Translation] - arXiv:2610.02254v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Interpretive Structural Modeling (ISM) is a well-known process for multi-criteria decision making.\u003c/li\u003e\n\u003cli\u003eThe success of ISM over other methodologies is its ability to model causal relationships, the binary scale of factors, and resulting hierarchical representation.\u003c/li\u003e\n\u003cli\u003eTraditionally, the modeling process is performed by repeated interactions with subject matter experts until consensus is reached.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02254v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Interpretive Structural Modeling (ISM) is a well-known process for multi-criteria decision making\u003c/li\u003e\n\u003cli\u003eThe success of ISM over other methodologies is its ability to model causal relationships, the binary scale of factors, and resulting hierarchical representation\u003c/li\u003e\n\u003cli\u003eTraditionally, the modeling process is performed by repeated interactions with subject matter experts until consensus is reached\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02255\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eMACTS-EM: Multi-Agent Collaborative Time Series Forecasting with Emergent Memory\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Pending Translation] - arXiv:2610.02255v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Time series forecasting remains a critical challenge across numerous domains.\u003c/li\u003e\n\u003cli\u003eDespite significant advancements, existing approaches struggle with complex phenomena such as regime shifts, cross-domain knowledge transfer, and multimodal data integration.\u003c/li\u003e\n\u003cli\u003eThis paper introduces Multi-Agent Collaborative Time Series Forecasting with Emergent Memory (MACTS-EM), a novel framework where specialised agents collaborate to achieve superior forecasting performance.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02255v1 Announce Type: new\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2610.02256\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eTRACE: A Reproducible Benchmark for Electricity Price Forecasting with Official Operational Text\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-10-05 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: [Translation Pending] - arXiv:2610.02256v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Electricity price forecasting (EPF) supports scheduling, bidding, and risk management in electricity markets, yet existing benchmarks focus mainly on numerical inputs, leaving the forecasting value of forecast-time textual context insufficiently evaluated.\u003c/li\u003e\n\u003cli\u003eWe introduce TRACE, a reproducible benchmark of 7,300 zone\u0026ndash;day instances pairing prices from five zones in a major U.S.\u003c/li\u003e\n\u003cli\u003emarket with official operational text available at the forecast cutoff.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2610.02256v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Electricity price forecasting (EPF) supports scheduling, bidding, and risk management in electricity markets, yet existing benchmarks focus mainly on…\u003c/li\u003e\n\u003cli\u003eWe introduce TRACE, a reproducible benchmark of 7,300 zone\u0026ndash;day instances pairing prices from five zones in a major U.S\u003c/li\u003e\n\u003cli\u003emarket with official operational text available at the forecast cutoff\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n",
  "wordCount": 6539,
  "readingTime": 31,
  "tableOfContents": "\u003cnav id=\"TableOfContents\"\u003e\n  \u003cul\u003e\n    \u003cli\u003e\u003ca href=\"#-in-depth-guide-to-this-issues-watch-list\"\u003e📖 In-depth Guide to This Issue\u0026rsquo;s Watch List\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-quick-takes-on-ai-hot-topics-from-x\"\u003e🌐 Quick Takes on AI Hot Topics from X\u003c/a\u003e\n      \u003cul\u003e\n        \u003cli\u003e\u003ca href=\"#topic-1-ai-video-of-jensen-huang-as-michael-jackson-wows-crowds\"\u003eTopic 1: AI Video of Jensen Huang as Michael Jackson Wows Crowds\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-2-openais-dots-ai-agents-launch-with-daily-improvement-pledge\"\u003eTopic 2: OpenAI\u0026rsquo;s Dots AI Agents Launch with Daily Improvement Pledge\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-3-trump-launches-super-intelligence-force-after-tech-accord\"\u003eTopic 3: Trump Launches Super Intelligence Force After Tech Accord\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-4-openai-launches-28-day-codex-improvement-sprint\"\u003eTopic 4: OpenAI Launches 28-Day Codex Improvement Sprint\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-5-tesla-expands-cybercab-robotaxi-fleet-in-texas-with-safety-first\"\u003eTopic 5: Tesla Expands Cybercab Robotaxi Fleet in Texas with Safety First\u003c/a\u003e\u003c/li\u003e\n      \u003c/ul\u003e\n    \u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-influencer-insights\"\u003e💡 Influencer Insights\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-appendix-todays-watch-list-update-source-list\"\u003e📚 Appendix: Today\u0026rsquo;s Watch List Update Source List\u003c/a\u003e\n      \u003cul\u003e\n        \u003cli\u003e\u003ca href=\"#stratechery-by-ben-thompson-a_full\"\u003eStratechery by Ben Thompson (A_full)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#openai-blog-a_full\"\u003eOpenAI Blog (A_full)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#two-minute-papers-b_introsearch\"\u003eTwo Minute Papers (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#arxiv-csai-b_introsearch\"\u003eArXiv cs.AI (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#arxiv-cscl-b_introsearch\"\u003eArXiv cs.CL (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#arxiv-cslg-b_introsearch\"\u003eArXiv cs.LG (B_intro+search)\u003c/a\u003e\u003c/li\u003e\n      \u003c/ul\u003e\n    \u003c/li\u003e\n  \u003c/ul\u003e\n\u003c/nav\u003e",
  "isDraft": false
}