{
  "title": "2026-06-17 AI Daily Update | AI Agents Begin Entering Real Workflows: From Phone Operations to Physical Experiments",
  "url": "https://miaok.ong/en/ai-daily/ai-daily-2026-06-17/",
  "date": "2026-06-17T07:00:00+08:00",
  "lastmod": "2026-06-17T07:00:00+08:00",
  "type": "ai-daily",
  "kind": "page",
  "language": "en",
  "description": "Today\u0026rsquo;s main theme is the progression of AI Agents from demos to real-world workflows: mobile agent evaluations, end-to-end development plugins, and hosted intelligent agents are all strengthening their execution capabilities. At the same time, the industry is starting to place more emphasis on reliable reasoning, testing, and contracts, focusing on how model answers can be verified. On-device and specialized small models also continue to gain traction, showing that efficiency, cost, and local deployment are becoming new points of competition.",
  "keywords": null,
  "tags": [],
  "categories": [],
  "author": "Mark (Miao) Kong",
  "image": "https://miaok.ong/images/avatar.jpg",
  "content": "\u003ch1 id=\"2026-06-17-ai-daily--ai-agents-begin-to-enter-real-workflows-from-mobile-operations-to-physical-experiments\"\u003e\n  2026-06-17 AI Daily | AI Agents Begin to Enter Real Workflows: From Mobile Operations to Physical Experiments\n  \u003ca class=\"heading-link\" href=\"#2026-06-17-ai-daily--ai-agents-begin-to-enter-real-workflows-from-mobile-operations-to-physical-experiments\"\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\u003eToday\u0026rsquo;s main theme is the transition of AI Agents from demos to real-world workflows: mobile agent evaluations, end-to-end development plugins, and hosted intelligent agents are all strengthening execution capabilities. At the same time, the industry is placing a greater emphasis on reliable reasoning, testing, and contracts, focusing on how model answers can be verified. On-device models and specialized small models also continue to gain traction, indicating that efficiency, cost, and local deployment are becoming new competitive battlegrounds.\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\u003eThe most noteworthy topic today is \u0026ldquo;AI Agents transitioning from demos to real-world workflows.\u0026rdquo; PhoneHarness redefines mobile agent evaluation, emphasizing the mixed use of GUI, CLI, and tools. Meanwhile, research on Web Agents reminds us that memory and skill modules are not free—they must be re-evaluated within the token budget.\u003c/p\u003e\n\u003cp\u003eThe second main theme is \u0026ldquo;reliable reasoning and interpretability.\u0026rdquo; CoRA investigates whether confidence scores are genuinely supported by the reasoning process, while papers on Lean 4 automatic formalization and the definition of LLM explanations are all asking the same question: How can model answers be verified and trusted?\u003c/p\u003e\n\u003cp\u003eAdditionally, the AI planning prototype from DeepMind and the UK government is worth reading for product and policy teams, as it shows AI entering high-friction public infrastructure scenarios like housing approvals. Context Compression, multilingual tokenizers, and Nemotron 3 Ultra can serve as technical supplements for improving model efficiency.\u003c/p\u003e\n\u003ch2 id=\"-ai-hot-topics-on-x\"\u003e\n  🌐 AI Hot Topics on X\n  \u003ca class=\"heading-link\" href=\"#-ai-hot-topics-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/h2\u003e\n\u003ch3 id=\"topic-1-spacex-acquires-cursor-maker-anysphere-in-60-billion-stock-deal\"\u003e\n  Topic 1: SpaceX Acquires Cursor Maker Anysphere in $60 Billion Stock Deal\n  \u003ca class=\"heading-link\" href=\"#topic-1-spacex-acquires-cursor-maker-anysphere-in-60-billion-stock-deal\"\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 Time: 13 hours ago, Related Posts: 78,000\u003c/li\u003e\n\u003cli\u003eWhat happened: According to trending news on X, SpaceX is acquiring Anysphere, the developer of the AI coding tool Cursor, in a $60 billion stock deal.\u003c/li\u003e\n\u003cli\u003eWhy it\u0026rsquo;s important: If true, this would indicate that aerospace and hard-tech companies are accelerating the integration of AI programming capabilities, highlighting the strategic value of code generation tools in engineering R\u0026amp;D, automated development, and enterprise productivity.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: Discussions on X are focused on the authenticity of the deal, whether the $60 billion valuation is excessive, SpaceX\u0026rsquo;s strategic intentions for acquiring an AI development tool, and whether Cursor will be deeply integrated into aerospace software and internal engineering workflows.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-2-matt-shumer-seeks-mobile-control-for-claude-code-ai-agent\"\u003e\n  Topic 2: Matt Shumer Seeks Mobile Control for Claude Code AI Agent\n  \u003ca class=\"heading-link\" href=\"#topic-2-matt-shumer-seeks-mobile-control-for-claude-code-ai-agent\"\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 Time: , Related Posts: 33\u003c/li\u003e\n\u003cli\u003eAbstract: Matt Shumer Seeks Mobile Control for Claude Code AI Agent:\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-3-openai-fixes-codex-outage-and-promises-rate-limit-resets\"\u003e\n  Topic 3: OpenAI Fixes Codex Outage and Promises Rate Limit Resets\n  \u003ca class=\"heading-link\" href=\"#topic-3-openai-fixes-codex-outage-and-promises-rate-limit-resets\"\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 Time: 11 hours ago, Related Posts: 2,400\u003c/li\u003e\n\u003cli\u003eWhat happened: OpenAI has fixed the Codex service outage and stated that it will reset the relevant rate limits for affected users.\u003c/li\u003e\n\u003cli\u003eWhy it\u0026rsquo;s important: Codex is a critical gateway for developers using AI programming tools. This outage highlights the crucial impact of AI infrastructure stability, availability, and quota management on the adoption of productivity tools.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: Discussions on X are mainly focused on the impact of the service interruption on development workflows, whether OpenAI\u0026rsquo;s compensation measures are adequate, and concerns from heavy users about the transparency and reliability of rate limits.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-4-openai-launches-developers-plugin-for-codex-coding-app\"\u003e\n  Topic 4: OpenAI Launches Developers Plugin for Codex Coding App\n  \u003ca class=\"heading-link\" href=\"#topic-4-openai-launches-developers-plugin-for-codex-coding-app\"\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 Time: 23 hours ago, Related Posts: 294\u003c/li\u003e\n\u003cli\u003eWhat happened: OpenAI has launched a developer plugin for its Codex coding application, which allows users to build, preview, and test iOS apps within Codex, optimizing the development process with SwiftUI Preview and hot reloading.\u003c/li\u003e\n\u003cli\u003eWhy it\u0026rsquo;s important: This shows that AI programming tools are evolving from code generation to end-to-end development environments, capable of handling writing, running, debugging, and feedback in a single workflow. This enhances the ability of AI agents to participate in real software development.\u003c/li\u003e\n\u003cli\u003eDiscussion summary: Discussions on X center on the competition between Codex and other tools like Claude Code and Cursor, and whether a plugin ecosystem will become a key moat for AI programming products. Supporters believe it streamlines the development process, while critics are concerned about stability, controllability, and practical efficiency in complex projects.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-5-ai-builders-embrace-agentic-loops-for-self-reliant-tasks\"\u003e\n  Topic 5: AI Builders Embrace Agentic Loops for Self-Reliant Tasks\n  \u003ca class=\"heading-link\" href=\"#topic-5-ai-builders-embrace-agentic-loops-for-self-reliant-tasks\"\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 Time: 6 hours ago, Related Posts: 105\u003c/li\u003e\n\u003cli\u003eWhat happened: AI developers are increasingly adopting \u0026ldquo;agentic loops\u0026rdquo; to build AI systems that can autonomously plan, execute, check, and iterate on tasks.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eWhy it matters:\u003c/strong\u003e This marks a shift in AI applications from single Q\u0026amp;A sessions to more autonomous workflows, promising to enhance efficiency in complex task processing, software development, data analysis, and automated operations.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eDiscussion landscape:\u003c/strong\u003e The discussion on X centers on whether agentic loops are truly reliable. Supporters believe they can reduce manual intervention and make AI assistants more practical, while skeptics worry about error accumulation, runaway costs, security boundaries, and the immaturity of evaluation standards.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch3 id=\"topic-6-anthropics-claude-managed-agents-speed-up-ai-production-deployment\"\u003e\n  Topic 6: Anthropic\u0026rsquo;s Claude Managed Agents Speed Up AI Production Deployment\n  \u003ca class=\"heading-link\" href=\"#topic-6-anthropics-claude-managed-agents-speed-up-ai-production-deployment\"\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\u003eCategory:\u003c/strong\u003e AI · News\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eOverview:\u003c/strong\u003e Trending since: 2 hours ago, Related posts: 217\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eWhat it is:\u003c/strong\u003e Anthropic has launched managed agents and a toolchain for enterprises and developers, built around Claude, to accelerate the transition of AI applications from prototype to production.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eWhy it matters:\u003c/strong\u003e This indicates that the competition among large models is shifting from pure model capabilities to enterprise-grade integration, automated deployment, and developer ecosystems. This could impact the adoption speed of AI in e-commerce, engineering, and office scenarios.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eDiscussion landscape:\u003c/strong\u003e The main topics on X are whether Anthropic is catching up to or even surpassing OpenAI, the practical value of Claude\u0026rsquo;s managed agents for enterprise AI deployment, and whether ecosystem collaborations with partners like Shopify will create new development workflows. The point of disagreement is whether these tools represent a productivity leap or are being overhyped by marketing and the \u0026ldquo;vibe coding\u0026rdquo; trend.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch4 id=\"ai-public-opinion-summary-on-x-today\"\u003e\n  AI Public Opinion Summary on X Today\n  \u003ca class=\"heading-link\" href=\"#ai-public-opinion-summary-on-x-today\"\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 narrative focuses on the rapid evolution of AI programming tools from \u0026ldquo;code assistants\u0026rdquo; to \u0026ldquo;end-to-end development and agentic execution platforms.\u0026rdquo; Whether it\u0026rsquo;s the rumor of Cursor\u0026rsquo;s high-priced acquisition by SpaceX, Codex\u0026rsquo;s launch of an iOS development plugin, or Anthropic\u0026rsquo;s release of managed agents, all signs point to the developer ecosystem and engineering workflows becoming the new battleground for large model competition. The consensus is that capabilities for code generation, preview, testing, deployment, and autonomous iteration are now considered strategic infrastructure for enterprise productivity and hard-tech R\u0026amp;D, making toolchain integration as important as the models themselves. The main disagreement lies in valuation versus actual utility: supporters believe these products will significantly shorten development cycles and drive AI assistants into production environments. Skeptics, however, think that acquisition rumors, plugin ecosystems, and \u0026ldquo;agentic loops\u0026rdquo; might be over-inflated by capital narratives and the \u0026ldquo;vibe coding\u0026rdquo; trend. Potential risks are concentrated in three areas: infrastructure stability and rate limits directly impact development workflows; autonomous agent execution could lead to error accumulation, runaway costs, and security boundary issues; and once a platform ecosystem becomes highly locked-in, it could diminish developers\u0026rsquo; control and freedom to migrate their toolchains.\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\u003ch1 id=\"analysis-of-ai-industry-dynamics-mid-june-2026\"\u003e\n  Analysis of AI Industry Dynamics (Mid-June 2026)\n  \u003ca class=\"heading-link\" href=\"#analysis-of-ai-industry-dynamics-mid-june-2026\"\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\u003ch2 id=\"1-todays-hot-topics-local-on-device-models-and-the-expansion-of-programming-agents-into-the-physical-world\"\u003e\n  1. Today\u0026rsquo;s Hot Topics: Local On-Device Models and the Expansion of Programming Agents into the Physical World\n  \u003ca class=\"heading-link\" href=\"#1-todays-hot-topics-local-on-device-models-and-the-expansion-of-programming-agents-into-the-physical-world\"\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\u003eThe focus of today\u0026rsquo;s discussion among influencers has undergone a significant \u0026ldquo;gravity shift\u0026rdquo;: \u003cstrong\u003efrom cloud-based super models to local on-device deployment, and from purely digital programming to physical world manipulation.\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eOn-device models hit the \u0026ldquo;sweet spot\u0026rdquo;:\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eAfter intensive testing, @zhixianio concluded that \u003cstrong\u003eQwen3.6-35B-A3B\u003c/strong\u003e running locally on a Mac has secured the \u0026ldquo;sweet spot\u0026rdquo; throne, surpassing remote LLMs in speed and intelligence, with a native multimodal experience that is even better than cloud-based large models.\u003c/li\u003e\n\u003cli\u003eRegarding Google\u0026rsquo;s newly released \u003cstrong\u003eGemma 4 12B Coder\u003c/strong\u003e, @zhixianio noted that despite optimizations, its 12B size remains a bottleneck for complex, \u0026ldquo;long-form, stateful, single-shot\u0026rdquo; generation tasks (like writing a complete game), showing a significant gap compared to 35B MoE models. He also tested Gemma 4\u0026rsquo;s audio capabilities, pointing out poor Chinese recognition but good performance in Japanese and English, and suggested that Quantization Aware Training (QAT) is a new approach to improving on-device efficiency.\u003c/li\u003e\n\u003cli\u003e@AI_Jasonyu corroborated the \u0026ldquo;on-device intelligence\u0026rdquo; trend from another angle: Baidu\u0026rsquo;s \u003cstrong\u003ePP-OCRv6\u003c/strong\u003e, with extremely few parameters (1.5MB), achieves higher OCR accuracy in the browser than large models like GPT-5.5, proving the huge advantage of \u0026ldquo;small models mastering vertical scenarios.\u0026rdquo;\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eAI Agents Enter the Physical World (AutoResearch):\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e@dotey highlighted NVIDIA GEAR lab\u0026rsquo;s \u003cstrong\u003eENPIRE\u003c/strong\u003e project. This is the first time a fully autonomous research loop for an AI programming agent (design, experiment, failure analysis, code iteration) has been implemented in a \u003cstrong\u003ereal physical environment\u003c/strong\u003e. The agent can autonomously control robots to perform high-precision tasks and discovered a \u0026ldquo;physical scaling law\u0026rdquo;: parallel robots can accelerate research. This marks a leap for AI capabilities from digital code generation to physical-world productivity.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2 id=\"2-unique-perspectives-and-industry-outlook\"\u003e\n  2. Unique Perspectives and Industry Outlook\n  \u003ca class=\"heading-link\" href=\"#2-unique-perspectives-and-industry-outlook\"\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\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eThe Real Moat in AI Programming: \u0026ldquo;Contracts\u0026rdquo; and \u0026ldquo;Tests,\u0026rdquo; Not Code Logic\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e@Pluvio9yte proposed that the secret to Vibe Coding lies in \u003cstrong\u003e\u0026ldquo;Contract First.\u0026rdquo;\u003c/strong\u003e Based on practical experience, he concluded that only by externalizing and clearly defining the contracts for APIs and data models can human-AI collaboration avoid context drift. This framework is more critical than just requirements or code alone.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e@ruanyf cites the case of a Cloudflare engineer replicating Next.js, sharply pointing out: \u003cstrong\u003e\u0026ldquo;Code itself no longer has a moat; testing is the new moat.\u0026rdquo;\u003c/strong\u003e This is because AI can easily replicate large projects, but the core barrier is the ability to pass high-quality tests and ensure stable operation.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eA Rational Look and \u0026ldquo;Contrarian View\u0026rdquo; on Claude Fable 5\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e@Pluvio9yte provided a deep-dive \u0026ldquo;contrarian\u0026rdquo; experience, differing from the hype: Fable 5 is extremely slow, token consumption isn\u0026rsquo;t as outrageous as imagined (about 1.5 times that of Opus), and while its capability boundaries are wider, it hasn\u0026rsquo;t reached a \u0026ldquo;stunning\u0026rdquo; level. It feels more like a hybrid of Opus 4.6++ and GPT-5.5++, prompting a call for rational use.\u003c/li\u003e\n\u003cli\u003e@dotey pointed out the token consumption black hole in Claude Code\u0026rsquo;s new Dynamic Workflows, where a simple task consumed 1.3 million tokens.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eProduction Relations and Economics in the AI Era\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e@ruanyf raised a sharp sociological question: After AI significantly boosts efficiency, will employees get time off? If there are no raises and no breaks, what is the point of AI for employees? He also calculated that the cost of unlimitedly using top-tier models for AI programming already far exceeds human programmer salaries, suggesting that businesses will need to weigh the ROI of AI in the future.\u003c/li\u003e\n\u003cli\u003e@vista8 shared a forward-looking perspective from the CEO of Factory AI: In the future, the most valuable people will be engineers who can deliver end-to-end business results, not just those who write code. Furthermore, within three years, the median token expenditure per employee will equal their salary.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eAI Product Design and Traffic Strategies\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e@Pluvio9yte suggested a way to eliminate the \u0026ldquo;AI feel\u0026rdquo; from UIs: use \u003cstrong\u003eDESIGN.md\u003c/strong\u003e files from famous brands as a constraint for AI generation to enhance the quality.\u003c/li\u003e\n\u003cli\u003e@gefei55 shared practical SEO experience, stressing the importance of evolving with Google\u0026rsquo;s algorithm and warning that low-quality AIGC content will eventually face a backlash. He also shared a highly profitable domain investment story and the feasibility of a high-pricing strategy for SaaS overseas.\u003c/li\u003e\n\u003cli\u003e@AI_Jasonyu observed that the paywall competition in the AI video space has shifted from competing on features to competing on \u0026ldquo;the explanation of the credit system.\u0026rdquo; The \u0026ldquo;vertical scenario + long-form to short-form video\u0026rdquo; logic, like that of OpusClip, is best suited for independent developers.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2 id=\"3-recommended-tools--resources\"\u003e\n  3. Recommended Tools \u0026amp; Resources\n  \u003ca class=\"heading-link\" href=\"#3-recommended-tools--resources\"\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\u003ctable\u003e\n  \u003cthead\u003e\n      \u003ctr\u003e\n          \u003cth style=\"text-align: left\"\u003eTool/Resource\u003c/th\u003e\n          \u003cth style=\"text-align: left\"\u003eRecommended by\u003c/th\u003e\n          \u003cth style=\"text-align: left\"\u003eCore Highlights \u0026amp; Use Case\u003c/th\u003e\n      \u003c/tr\u003e\n  \u003c/thead\u003e\n  \u003ctbody\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003ebaoyu-design Skill\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@dotey\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eLocal design-to-code\u003c/strong\u003e. Supports importing Figma files, generating design systems and PPTs locally, and can even export to editable PPTX files.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003einfo-digest Skill\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@dotey\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eAI News Digesting Assistant\u003c/strong\u003e. Baoyu\u0026rsquo;s public daily writing Skill, containing adaptable strategies like reader perspective, fact-checking, and refined formatting.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003egetdesign.md\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@Pluvio9yte\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eThe ultimate tool to remove the \u0026ldquo;AI feel\u0026rdquo; from UI design\u003c/strong\u003e. A collection of DESIGN.md system files from real brands like Linear, Vercel, and Apple, which can be fed to an AI to generate high-quality UIs.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003ePapr\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@vista8\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eA lightweight, open-source RSS client\u003c/strong\u003e. Supports connecting your own API key for AI summaries and Q\u0026amp;A.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eFigma Chrome Extension\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@vista8\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eGame-changer for website cloning\u003c/strong\u003e. One-click conversion of any webpage element into editable layers and import into Figma.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003ePP-OCRv6\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@AI_Jasonyu\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eUltimate on-device OCR\u003c/strong\u003e. A 1.5MB model that can run in the browser, surpassing large models like GPT-5.5 in speed and accuracy. Fully open-source.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eApp Store Review Analysis Tool\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@vista8\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eOpen-source user feedback miner\u003c/strong\u003e. Can scrape reviews for any app and use an LLM to analyze pain points and opportunities.\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eGPT Image Prompt\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@dotey (from @Ciri_ai)\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003ePhoto to doodle illustration\u003c/strong\u003e. A specific prompt to transform photos into a \u0026ldquo;decorative folk flat doodle style.\u0026rdquo;\u003c/td\u003e\n      \u003c/tr\u003e\n      \u003ctr\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eMaccy / Mos\u003c/strong\u003e\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e@Pluvio9yte\u003c/td\u003e\n          \u003ctd style=\"text-align: left\"\u003e\u003cstrong\u003eMac productivity duo\u003c/strong\u003e. Maccy is an open-source clipboard tool, and Mos solves the unnatural scrolling direction issue with external mice.\u003c/td\u003e\n      \u003c/tr\u003e\n  \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2 id=\"-appendix-todays-watch-list-source-update\"\u003e\n  📚 Appendix: Today\u0026rsquo;s Watch List Source Update\n  \u003ca class=\"heading-link\" href=\"#-appendix-todays-watch-list-source-update\"\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; 22 sources covered; 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/fox-buys-roku-the-problem-with-foxs-smart-strategy-streaming-that-works/\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eFox Buys Roku, The Problem With Fox’s Smart Strategy, Streaming That Works\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 18:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - The market hates Fox\u0026rsquo;s acquisition of Roku, but the company is trading extraction from rights holders for leverage as a renter.\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003e$15\u003c/strong\u003e/month \u003cem\u003eor\u003c/em\u003e \u003cem\u003e\u003cstrong\u003e$150\u003c/strong\u003e\u003c/em\u003e/year.\u003c/li\u003e\n\u003cli\u003eSubstantive analysis of the day\u0026rsquo;s news via three emails or a podcast per week.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eStrategy Interviews\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003eInterviews with leading public company CEOs, private company founders, and discussions with fellow analysts.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003eThe market hates Fox\u0026rsquo;s acquisition of Roku, but the company is trading extraction from rights holders for leverage as a renter.\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\u003cstrong\u003e\u003ca href=\"https://openai.com/index/deployment-simulation\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePredicting model behavior before release by simulating deployment\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 08:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - Before releasing new models, labs need to understand not only what they can do, but also how they will behave in real-world use, including where they might introduce new risks.\n\u003cul\u003e\n\u003cli\u003eThis becomes even more important as capabilities increase.\u003c/li\u003e\n\u003cli\u003eAs part of our pre-deployment safety review, we use targeted evaluations, red teaming, and other checks to understand model behavior.\u003c/li\u003e\n\u003cli\u003eWe are now beginning to use a method to simulate model deployments before they happen, which adds a complementary signal: a deployment-like preview of a candidate model\u0026rsquo;s behavior before it reaches users.\u003c/li\u003e\n\u003cli\u003eDeployment simulation is a method of simulating future deployments before they occur.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003eOpenAI introduces Deployment Simulation, a method to predict AI model behavior before deployment using real conversation data to improve safety and evaluation a…\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=\"google-deepmind-blog-a_full\"\u003e\n  Google DeepMind Blog (A_full)\n  \u003ca class=\"heading-link\" href=\"#google-deepmind-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\u003cstrong\u003e\u003ca href=\"https://deepmind.google/blog/unlocking-uk-house-building-with-ai-accelerated-planning/\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eUnlocking UK house-building with AI-accelerated planning\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-17 05:29 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - The UK government is partnering with Google DeepMind to build a new AI prototype aimed at making faster housing decisions.\n\u003cul\u003e\n\u003cli\u003eThis article from the Google DeepMind Blog explains how to unlock UK house-building with AI-accelerated planning, shaping the broader AI and infrastructure landscape.\u003c/li\u003e\n\u003cli\u003eAfter unlocking UK house-building through AI-accelerated planning, it also brings practical implications for founders, operators, and investors.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003eUK government partners with Google DeepMind to build a new AI-powered prototype aimed at faster housing decisions.\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=l72ufA-4SzE\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eThey Looked Inside Claude’s AI\u0026rsquo;s Mind. It Got Weird\u003c/a\u003e\u003c/strong\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 23:53 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - ❤️ Check out Lambda and sign up for their GPU Cloud here:.\n\u003cul\u003e\n\u003cli\u003e📝 The paper is available here:.\u003c/li\u003e\n\u003cli\u003eAdam Bridges, Benji Rabhan, B Shang, Cameron Navor, Charles Ian Norman Venn, Christian Ahlin, Eric T, Fred R, Gordon Child, Juan Benet, Michael Tedder, Owen Skarpness, Richard Sundvall, Ryan Stankye, Shawn Becker, Steef, Taras Bobrovytsky, Tazaur Sagenclaw, Tybie Fitzhugh, Ueli Gallizzi.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eThey delved deep into the mind of Claude AI.\n\u003cul\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003e❤️ Check out Lambda here and sign up for their GPU Cloud:\u003c/li\u003e\n\u003cli\u003e📝 The paper is available here:\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\u003cli\u003eAdam Bridges, Benji Rabhan, B Shang, Cameron Navor, Charles Ian Norman Venn, Christian Ahlin, Eric T, Fred R, Gordon Child, Juan Benet, Michael Tedder, Owen Ska…\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-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/2606.14838\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eA Definition of Good Explanations and the Challenges Explaining LLM Outputs\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14838v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: How to define a good explanation is a long-standing philosophical debate that has recently seen renewed interest in the context of AI outputs.\u003c/li\u003e\n\u003cli\u003eExplainability is crucial for the adoption of AI in many contexts, but in order to produce good explanations for AI systems, we must first have an understanding of what constitutes a good explanation.\u003c/li\u003e\n\u003cli\u003eIn this paper, we propose a definition inspired by the concept of counterfactual explanations, but we argue that one must also consider the interlocutor\u0026rsquo;s prior beliefs about each fact that might be provided in the explanation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14838v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: How to define a good explanation is a long-standing philosophical debate which has found recent renewed interest in the context of AI outputs\u003c/li\u003e\n\u003cli\u003eExplainability is crucial for AI adoption in many contexts, but in order to produce good explanations of AI systems, we must first have an understanding of what…\u003c/li\u003e\n\u003cli\u003eIn this paper we propose a definition inspired by the notion of counterfactual explanations, however we argue that one must also take into account the interlocu…\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/2606.14885\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eDr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace Expansion\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14885v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Agentic search over large corpora relies on retriever-mediated interfaces (e.g., BM25 or ColBERT) for scalable candidate discovery.\u003c/li\u003e\n\u003cli\u003eWhile effective at ranking relevant documents, these interfaces only expose evidence as ranked results or bounded document views, limiting an agent\u0026rsquo;s ability to reorganize materials and validate cross-document constraints.\u003c/li\u003e\n\u003cli\u003eDirect Corpus Interaction (DCI) addresses this limitation by exposing shell-executable corpus operations for flexible search, filtering, comparison, and validation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14885v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Agentic search over large corpora relies on retriever-mediated interfaces (e.g., BM25 or ColBERT) for scalable candidate discovery\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eWhile effective at ranking relevant documents, these interfaces expose evidence only as ranked results or bounded document views, limiting agents\u0026rsquo; ability to re…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eDirect Corpus Interaction (DCI) addresses this limitation by exposing shell-executable corpus operations for flexible search, filtering, comparison, and verific…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2606.14892\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eRelational Structural Causal Models\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14892v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: An artificial intelligence must have a causal model of its environment, supporting reasoning about interventions and counterfactuals, and also a compositional model of its environment, supporting generalization to unseen combinations of objects.\u003c/li\u003e\n\u003cli\u003eIn this work, we formally study when and how such a model can be learned.\u003c/li\u003e\n\u003cli\u003eWe develop relational structural causal models, extending structural causal models (Pearl 2009) to settings where objects and their relations vary.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14892v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: An artificial intelligence must have a model of its environment that is causal, supporting reasoning about interventions and counterfactuals, and also…\u003c/li\u003e\n\u003cli\u003eIn this work, we formally study when and how such a model can be learned\u003c/li\u003e\n\u003cli\u003eWe develop relational structural causal models, extending structural causal models (Pearl 2009) to settings where objects and their relations vary\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/2606.14923\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eTrust Between AI Agents: Measuring Formation, Breakage, and Recovery, with Implications for Governing Multi-Agent Systems\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14923v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: As language-model agents increasingly work in teams, each agent must decide how much to trust its teammates.\u003c/li\u003e\n\u003cli\u003eHowever, we lack a standard way to measure trust between AI agents.\u003c/li\u003e\n\u003cli\u003eWe propose a behavioral measure based on costly verification.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14923v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: As language-model agents increasingly work in teams, each agent must decide how much to trust its teammates\u003c/li\u003e\n\u003cli\u003eYet we lack a standard way to measure trust between AI agents\u003c/li\u003e\n\u003cli\u003eWe propose a behavioral measure based on costly verification\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/2606.14935\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePrologMCP: A Standardized Prolog Tool Interface for LLM Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14935v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: State-of-the-art language models fine-tuned for reasoning still fail on deep deductive tasks, and the cost of improving performance by expanding internal reasoning is also poor.\u003c/li\u003e\n\u003cli\u003eSymbolic delegation offers a complementary path: the language model translates the problem, while the solver performs the reasoning.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eHowever, current autoformalization pipelines for logic programming are typically bespoke integrations tied to particular tasks or agents.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14935v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Frontier reasoning-tuned language models still fail on deductive tasks at depth, and the cost of improved performance through extended internal reason…\u003c/li\u003e\n\u003cli\u003eSymbolic delegation offers a complementary route: a language model translates the problem, while a solver performs the inference\u003c/li\u003e\n\u003cli\u003eHowever, current autoformalization pipelines for logic programming are typically bespoke integrations tied to particular tasks or agents\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/2606.14941\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSemantics-Enhanced Retrieval-Augmented Time Series Forecasting\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14941v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Time series forecasting models often benefit from historical patterns.\u003c/li\u003e\n\u003cli\u003eInspired by Retrieval-Augmented Generation (RAG), recent research explored retrieving relevant historical time series segments to enhance forecasting.\u003c/li\u003e\n\u003cli\u003eHowever, relying solely on time series similarity is often insufficient for retrieval under non-stationarity.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14941v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Time series forecasting models often benefit from historical patterns\u003c/li\u003e\n\u003cli\u003eInspired by Retrieval-Augmented Generation (RAG), recent research explored retrieving relevant historical time series segments to enhance forecasting\u003c/li\u003e\n\u003cli\u003eHowever, relying solely on time series similarity is often insufficient for retrieval under non-stationarity\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/2606.14997\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eAI Engram: In Search of Memory Traces in Artificial Intelligence\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14997v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Memory formation is fundamental to intelligence, yet whether deep neural networks preserve identifiable memory traces analogous to biological memory units remains an open question.\u003c/li\u003e\n\u003cli\u003eThis work introduces a geometric framework to identify such \u0026ldquo;AI engrams\u0026rdquo; by formalizing the neuroscientific criteria of specificity, reactivation, sufficiency, and necessity as a constrained inverse problem.\u003c/li\u003e\n\u003cli\u003eWe derive a closed-form estimator that isolates individual memory traces from globally entangled parameters and show that this biologically-derived solution corresponds to a natural gradient update on the parameter manifold.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14997v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Memory formation is fundamental to intelligence, yet whether deep neural networks preserve identifiable memory traces analogous to biological memory u…\u003c/li\u003e\n\u003cli\u003eThis work introduces a geometric framework to identify such \u0026ldquo;AI engrams\u0026rdquo; by formalizing the neuroscientific criteria of specificity, reactivation, sufficiency,…\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 derive a closed-form estimator that isolates individual memory traces from globally entangled parameters, and show that this biologically-derived solution co…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2606.15029\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eMetric Match: A Subset Selection Approach to Evaluating LLM Judge Reliability\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15029v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: LLM judges are used to reduce the need for expensive human labor when evaluating open-ended text generation.\u003c/li\u003e\n\u003cli\u003eHowever, the reliability of these judges largely depends on their consistency with human raters—a property that itself relies on costly human annotations.\u003c/li\u003e\n\u003cli\u003eIn this work, we develop a method (Metric Match) for estimating correlation-based reliability metrics of LLM judges from limited annotations.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15029v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: LLM judges are used to reduce the need for costly human labor in evaluating open-ended text generation\u003c/li\u003e\n\u003cli\u003eHowever, the reliability of these judges depends critically on their alignment with human raters \u0026ndash; a property that itself depends on costly human annotations\u003c/li\u003e\n\u003cli\u003eIn this work, we develop a method (Metric Match) for estimating correlation-based reliability metrics of LLM judges from limited annotations\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/2606.15034\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eOSGuard: A Benchmark for Safety in Computer-Use Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15034v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Computer-use agents are increasingly evaluated on whether they complete realistic desktop and Web tasks.\u003c/li\u003e\n\u003cli\u003eHowever, task success alone can miss failures where an agent reaches the nominal goal through an unsafe shortcut.\u003c/li\u003e\n\u003cli\u003eWe introduce OSGuard, a dual-granularity benchmark suite for evaluating safety in computer-use agents under benign, unchanged user instructions.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN 要点:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15034v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Computer-use agents are increasingly evaluated by whether they complete realistic desktop and web tasks\u003c/li\u003e\n\u003cli\u003eHowever, task success alone can miss failures in which an agent reaches the nominal goal through an unsafe shortcut\u003c/li\u003e\n\u003cli\u003eWe introduce OSGuard, a dual-granularity benchmark suite for evaluating safety in computer-use agents under benign, unchanged user instructions\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/2606.15038\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eFusion is not one-size-fits-all: Cross-Modal Representation Alignment for Time-to-Event Modeling\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15038v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Accurately predicting Time-to-Event (TTE) from multimodal clinical data remains challenging due to modal imbalances and distribution shifts.\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 a foundation model-driven framework for cross-modal representation alignment between CT imaging and longitudinal EHR data, designed to generalize across tasks and institutions.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eCT and EHR modalities are encoded independently using domain-specific foundation models and aligned in a shared latent space through four principled fusion strategies: late fusion, contrastive alignment, cross-attention, and co-attention.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eEN Highlights:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15038v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Accurate time-to-event (TTE) prediction from multimodal clinical data remains challenging due to modality imbalance and distribution shift\u003c/li\u003e\n\u003cli\u003eWe introduce a foundation model-driven framework for cross-modal representation alignment between CT imaging and longitudinal EHR data, designed to generalize a…\u003c/li\u003e\n\u003cli\u003eCT and EHR modalities are encoded independently using domain-specific foundation models and aligned in a shared latent space through four principled fusion stra…\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/2606.14832\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePhoneHarness: Harnessing Phone-Use Agents through Mixed GUI, CLI, and Tool Actions\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14832v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Phone agents are increasingly expected to complete real mobile workflows rather than merely predict the next screen action.\u003c/li\u003e\n\u003cli\u003eHowever, much of the current mobile-agent literature still evaluates agents primarily as GUI controllers that observe a screen, emit taps and swipes, and are scored based on the target application state.\u003c/li\u003e\n\u003cli\u003eReal phone-use tasks are broader: they require deciding when to use app GUIs, device-side commands, or structured tools, while leaving evidence that the intended side effects have indeed occurred.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14832v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Phone agents are increasingly expected to complete real mobile workflows rather than merely predict the next screen action\u003c/li\u003e\n\u003cli\u003eHowever, much of the current mobile-agent literature still evaluates agents primarily as GUI controllers that observe a screen, emit taps and swipes, and are sc…\u003c/li\u003e\n\u003cli\u003eReal phone-use tasks are broader: they require deciding when to use app GUIs, device-side commands, or structured tools, while leaving evidence that the intende…\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/2606.14867\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eEvaluating the Robustness of Proof Autoformalization in Lean 4\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14867v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Proof autoformalization aims to translate informal mathematical proofs written in natural language into formal proofs in a formal language (e.g., Lean 4).\u003c/li\u003e\n\u003cli\u003eSeveral works have developed LLM-based models for proof autoformalization.\u003c/li\u003e\n\u003cli\u003eHowever, existing evaluations often focus on translating well-formed informal proofs from curated datasets.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14867v1 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: Proof autoformalization aims to translate a mathematical informal proof written in natural language into a formal proof in a formal language such as L…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eSeveral works have developed LLM-based models for proof autoformalization\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eHowever, existing evaluations have typically focused on translating well-formed informal proofs from curated datasets\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2606.14875\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eContext Compression Is Not One Thing: Readable Symbolic Re-expression vs. Coherent Summary at Matched Budget\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14875v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: We study context compression for multi-hop question answering with small language models.\u003c/li\u003e\n\u003cli\u003eWe propose Telegraph English, a readable symbolic format that rewrites retrieved passages into structured entity-relation statements, preserving reasoning evidence at a lower token cost.\u003c/li\u003e\n\u003cli\u003eIn controlled experiments on MuSiQue, TwoWiki, and HotpotQA, Telegraph English outperforms three matched-budget compression baselines (character-level deletion, truncation, and random subsampling) with gains of 13 to 20 F1 percentage points.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14875v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: We study context compression for multi-hop question answering with small language models\u003c/li\u003e\n\u003cli\u003eWe propose Telegraph English, a readable symbolic format that rewrites retrieved passages into structured entity-relation statements, preserving reasoning evide…\u003c/li\u003e\n\u003cli\u003eIn controlled experiments on MuSiQue, TwoWiki, and HotpotQA, Telegraph English outperforms three matched-budget compression baselines (character-level deletion,…\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/2606.14943\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSimplifying the Modeling of Arbitrary Conditionals in Natural Language\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublished: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14943v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Causal Transformers model sequences through an autoregressive factorization of the joint distribution, which enables efficient left-to-right decoding and conditional likelihood computation.\u003c/li\u003e\n\u003cli\u003eHowever, they cannot tractably sample from or evaluate arbitrary conditionals \u0026ndash; e.g., a block of text conditioned on past and future tokens.\u003c/li\u003e\n\u003cli\u003eRecent work has aimed to solve this with novel architectures, but they often result in suboptimal modeling of such conditionals and degenerate generation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14943v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Causal Transformers model sequences through an autoregressive factorization of the joint distribution, which enables efficient left-to-right decoding…\u003c/li\u003e\n\u003cli\u003eHowever, they cannot tractably sample from or evaluate arbitrary conditionals \u0026ndash; e.g., a block of text conditioned on past and future tokens\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/2606.14961\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eCoRA: Confidence-Rationale Alignment for Reliable Chain-of-Thought Reasoning\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14961v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Chain-of-thought (CoT) reasoning can improve LLM performance, but high answer confidence can be misleading when the accompanying CoT rationale is seemingly plausible but incomplete or unsupported.\u003c/li\u003e\n\u003cli\u003eWe study confidence-rationale alignment: whether a model\u0026rsquo;s confidence in its committed answer is justified by its generated rationale.\u003c/li\u003e\n\u003cli\u003eWe introduce a GRPO-based reinforcement learning framework that jointly rewards answer correctness, committed-answer probability, and rubric-based rationale support, where the rubric assesses grounding, coherence, task-matching, and connection to the chosen answer without revealing the gold answer to the judge.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14961v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Chain-of-thought (CoT) reasoning can improve LLM performance, but high answer confidence may be misleading when the accompanying CoT rationale is plau…\u003c/li\u003e\n\u003cli\u003eWe study confidence\u0026ndash;rationale alignment: whether a model\u0026rsquo;s confidence in its committed answer is justified by its generated rationale\u003c/li\u003e\n\u003cli\u003eWe introduce a GRPO-based reinforcement learning framework that jointly rewards answer correctness, committed-answer probability, and rubric-based rationale sup…\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/2606.15007\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eNemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eRelease Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15007v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: We introduce Nemotron 3 Ultra, a Mixture-of-Experts hybrid Mamba-Attention language model with a total of 550 billion and 55 billion active parameters.\u003c/li\u003e\n\u003cli\u003eWe pre-trained Nemotron 3 Ultra on 20 trillion text tokens, then extended the context length to 1M tokens, and post-trained using Supervised Fine-Tuning (SFT), Reinforcement Learning (RL), and Multi-teacher On-Policy Distillation (MOPD).\u003c/li\u003e\n\u003cli\u003eNemotron 3 Ultra is our most powerful model to date, employing several key technologies - LatentMoE, Multi-Token Prediction (MTP), NVFP4 pre-training, multi-environment RLVR, MOPD, and inference budget control.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15007v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: We introduce Nemotron 3 Ultra, a 550 billion total and 55 billion active parameter Mixture-of-Experts Hybrid Mamba-Attention language model\u003c/li\u003e\n\u003cli\u003eWe pre-trained Nemotron 3 Ultra on 20 trillion text tokens, then extended the context length to 1M tokens, and post-trained using Supervised Fine Tuning (SFT),…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eNemotron 3 Ultra is our most capable model yet, employing multiple key technologies - LatentMoE, Multi Token Prediction (MTP), NVFP4 pre-training, multi-environ…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2606.15017\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eAre Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15017v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Online web agents often augment a base actor with memory, workflow, or skill modules.\u003c/li\u003e\n\u003cli\u003eThese modules can improve performance, but they also consume test-time tokens, a cost rarely reported alongside the actor\u0026rsquo;s inference cost.\u003c/li\u003e\n\u003cli\u003eWe study online augmentation, where this overhead is paid on every task, and re-evaluate its benefits under a fixed total inference budget.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15017v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Online web agents often augment a base actor with memory, workflow, or skill modules\u003c/li\u003e\n\u003cli\u003eThese modules can improve performance, but they also consume test-time tokens, a cost rarely reported alongside the actor\u0026rsquo;s inference cost\u003c/li\u003e\n\u003cli\u003eWe study online augmentation, where this overhead is paid on every task, and re-evaluate its benefits under a fixed total inference budget\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/2606.15026\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eDeep Temporal Modeling and Ensemble Fusion for Multimodal Emotion Recognition from Physiological Signals\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15026v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Physiological stress and emotion recognition are important for health monitoring and affective computing.\u003c/li\u003e\n\u003cli\u003eIn this work, we present a comprehensive evaluation of deep learning models such as Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), and Transformer on the WESAD dataset for multimodal emotion recognition using wrist and chest sensor signals.\u003c/li\u003e\n\u003cli\u003eWe perform ablation studies to assess the individual contributions of each modality by training models on wrist-only and chest-only inputs.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15026v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Physiological stress and emotion recognition are important for health monitoring and affective computing\u003c/li\u003e\n\u003cli\u003eIn this work, we present a comprehensive evaluation of deep learning models such as Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), and Tr…\u003c/li\u003e\n\u003cli\u003eWe perform ablation studies to assess the individual contributions of each modality by training models on wrist-only and chest-only inputs\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/2606.15037\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eReportQA: QA-Based Radiology Report Evaluation\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15037v1 Announce Type: new.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eAbstract: Radiology report evaluation is essential for advancing automated report generation.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eNatural language generation metrics have limited clinical relevance.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eClinical efficacy (CE) metrics evaluate important medical findings, but focus mainly on presence and cover only a limited set of entities.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eEN Highlights:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15037v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Radiology report evaluation is essential for advancing automated report generation\u003c/li\u003e\n\u003cli\u003eNatural language generation metrics have limited clinical relevance\u003c/li\u003e\n\u003cli\u003eClinical efficacy (CE) metrics evaluate important medical findings, but focus mainly on presence and cover only a limited set of entities\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/2606.15044\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eEquity with Efficiency: An Empirical Study of Tokenizers for Multilingual Large Language Models\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.15044v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Multilingual Large Language Models (LLMs) depend on subword tokenization to bridge discrete text and continuous neural representation.\u003c/li\u003e\n\u003cli\u003eState-of-the-art multilingual LLMs often use Byte-level Byte-Pair Encoding (BPE) tokenizers that structurally favor high-resource languages and Latin scripts.\u003c/li\u003e\n\u003cli\u003eFor speakers of underrepresented languages, particularly those across Southeast Asia, this bias inflates inference costs and widens the cross-lingual capability gap.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.15044v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Multilingual large language models (LLMs) depend on subword tokenization to bridge discrete text and continuous neural representation\u003c/li\u003e\n\u003cli\u003eState-of-the-art multilingual LLMs often use Byte-level Byte-Pair Encoding (BPE) tokenizers that structurally favor high-resource languages and Latin scripts\u003c/li\u003e\n\u003cli\u003eFor speakers of underrepresented languages, particularly those across Southeast Asia, this bias inflates inference costs and widens cross-lingual capability gap…\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-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/2606.14801\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eQPILOTS: Efficient Test-Time Q-Steering for Flow Policies\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14801v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Flow-matching and diffusion policies are expressive action generators, but optimizing them with temporal-difference reinforcement learning (RL) remains difficult.\u003c/li\u003e\n\u003cli\u003eEffective policy extraction requires leveraging the critic\u0026rsquo;s action-gradients, but directly backpropagating this signal through a multi-step denoising process can be numerically unstable.\u003c/li\u003e\n\u003cli\u003eExisting methods address this issue by discarding gradient information, distilling the policy into a simpler single-step actor, or repeatedly fine-tuning the denoising policy as the critic improves.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14801v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Flow-matching and diffusion policies are expressive action generators, but optimizing them with temporal-difference reinforcement learning (RL) remain…\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eEffective policy extraction requires exploiting the critic\u0026rsquo;s action gradient, yet directly backpropagating this signal through a multi-step denoising process ca…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eExisting methods work around this either by discarding gradient information, distilling the policy into a simpler one-step actor, or repeatedly fine-tuning the…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2606.14865\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eGRAPE: Guided Parameter-Space Evolution for Compact Adversarial Robustness\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14865v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Adversarial Training (AT) improves neural network robustness, but most methods train a fixed parameter space from the start.\u003c/li\u003e\n\u003cli\u003eThis paper asks whether the order in which parameters become optimizable can affect the final robust solution, even when the final architecture or computation budget is controlled.\u003c/li\u003e\n\u003cli\u003eWe propose GRAPE (Guided Parameter-Space Evolution), a training framework for compact adversarial robustness.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14865v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Adversarial Training (AT) improves neural network robustness, but most methods train a fixed parameter space from the start\u003c/li\u003e\n\u003cli\u003eThis paper asks whether the order in which parameters become optimizable can affect the final robust solution, even when the final architecture or computation b…\u003c/li\u003e\n\u003cli\u003eWe propose GRAPE, Guided Parameter-Space Evolution, a training framework for compact adversarial robustness\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/2606.14898\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003e{\\alpha}-Fair Insurance Pricing: A Fairness Continuum\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14898v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Fairness in insurance pricing remains a long-standing and deeply debated puzzle.\u003c/li\u003e\n\u003cli\u003eOn one hand, insurers, driven by profitability considerations, set premiums that differentiate across individual risks to achieve actuarial fairness.\u003c/li\u003e\n\u003cli\u003eOn the other hand, insurance serves a critical societal function by pooling risks across a population, incentivizing cross-subsidization among groups to promote solidarity fairness.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14898v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Fairness in insurance pricing remains a long-standing and deeply debated puzzle\u003c/li\u003e\n\u003cli\u003eOn one hand, insurers, driven by profitability considerations, set premiums that differentiate across individual risks to achieve actuarial fairness\u003c/li\u003e\n\u003cli\u003eOn the other hand, insurance serves a critical societal function by pooling risks across a population, motivating cross-subsidization among groups to promote so…\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/2606.14900\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eGRASP: Gradient-Aligned Sequential Parameter Transfer for Memory-Efficient Multi-Source Learning\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14900v1 Announcement Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Multi-source transfer learning faces a fundamental scalability bottleneck: existing approaches either require loading all K source models into memory simultaneously during parameter fusion, which needs O(K) memory, or deploying all models at inference, making production deployment infeasible.\u003c/li\u003e\n\u003cli\u003eWe propose GRASP (Gradient-Aligned Sequential Parameter Transfer), which achieves superior knowledge integration while maintaining O(1) memory consumption through three key innovations: (1) sequential processing, merging one source at a time into an evolving target model; (2) parameter-gradient alignment, selectively transferring only parameters whose optimization direction aligns with the target domain to avoid negative transfer; and (3) iterative fine-tuning to adapt the transferred knowledge before integrating the next source.\u003c/li\u003e\n\u003cli\u003eExtensive experiments across three continual learning benchmarks (Yearbook, CLEAR-10, CLEAR-100), spanning 10 to 108-year temporal distribution shifts and four architectures (1.3M to 25.6M parameters), show that GRASP achieves an average accuracy of 93.5% across all datasets and architectures, compared to 71.7% for ensemble methods, while requiring only constant memory, whereas standard multi-source fusion for K models requires memory.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14900v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Multi-source transfer learning faces a fundamental scalability bottleneck: existing approaches require either loading all K source models into memory…\u003c/li\u003e\n\u003cli\u003eWe propose GRASP (Gradient-Aligned Sequential Parameter Transfer), which achieves superior knowledge integration while maintaining O(1) memory consumption throu…\u003c/li\u003e\n\u003cli\u003eExtensive experiments across three continual learning benchmarks (Yearbook, CLEAR-10, CLEAR-100) spanning 10 to 108-year temporal distribution shifts and four a…\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/2606.14929\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003ePolicy Regret for Embedding Model Routing: Contextual Bandits with Low-Rank Experts\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14929v1 Announcement Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: Modern recommendation systems increasingly rely on dynamically routing diverse queries to multiple embedding models.\u003c/li\u003e\n\u003cli\u003eDespite its practical significance, this problem remains poorly understood under realistic conditions like adversarial queries, bandit feedback, and limited observability of the models.\u003c/li\u003e\n\u003cli\u003eWe formalize embedding model routing as an adversarial contextual linear bandit with low-rank experts, where contexts are queries, actions are items, and experts are the embedding models that operate on a low-rank latent representation space.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14929v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Modern recommendation systems increasingly rely on dynamically routing diverse queries to multiple embedding models\u003c/li\u003e\n\u003cli\u003eDespite its practical significance, this problem remains poorly understood under realistic conditions like adversarial queries, bandit feedback, and limited obs…\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 formalize embedding model routing as an adversarial contextual linear bandit with low-rank experts, where contexts are queries, actions are items, and expert…\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003e\u003ca href=\"https://arxiv.org/abs/2606.14934\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eSeparable Neural Architectures as Physical World Models: from Mathematical Theory to Applications\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePosted: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14934v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: This work introduces the Separable Neural Architecture (SNA), a function representational class combining neural approximation with tensor decomposition.\u003c/li\u003e\n\u003cli\u003eThe SNA decouples localized coordinate functions (atoms) from global interactions governed by a sparse, low-rank interaction object.\u003c/li\u003e\n\u003cli\u003eThis architecture possesses a compact and smooth inductive bias well-suited for solving partial differential equations (PDEs).\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14934v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: This work introduces the Separable Neural Architecture (SNA), a function representational class combining neural approximation with tensor decompositi…\u003c/li\u003e\n\u003cli\u003eThe SNA decouples localized coordinate functions (atoms) from global interactions governed by a sparse, low-rank interaction object\u003c/li\u003e\n\u003cli\u003eThis architecture possesses a compact and smooth inductive bias well-suited for solving partial differential equations (PDEs)\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/2606.14945\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eRemember, Don\u0026rsquo;t Re-read: Stateful ReAct Agents for Token-Efficient Autonomous Experimentation\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePosted: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14945v1 Announce Type: new.\n\u003cul\u003e\n\u003cli\u003eAbstract: The autoresearch pattern enables autonomous experimentation by having a large language model (LLM) iteratively modify code to optimize a target metric.\u003c/li\u003e\n\u003cli\u003eIts stateless design, however, reconstructs experimental context from scratch at every iteration, incurring $O(n)$ token cost per iteration and $O(n^{2})$ total.\u003c/li\u003e\n\u003cli\u003eThis work reformulates the pattern as a stateful ReAct agent using LangGraph, where typed persistent state carries experimental history across iterations via a tool-call interface.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Highlights:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14945v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: The autoresearch pattern enables autonomous experimentation by having a large language model (LLM) iteratively modify code to optimize a target metric\u003c/li\u003e\n\u003cli\u003eIts stateless design, however, reconstructs experimental context from scratch at every iteration, incurring $O(n)$ token cost per iteration and $O(n^{2})$ total\u003c/li\u003e\n\u003cli\u003eThis work reformulates the pattern as a stateful ReAct agent using LangGraph, where typed persistent state carries experimental history across iterations via a…\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/2606.14956\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eA Comparative Study of Graph Neural Network Layer Selection for Interaction Modelling in Driving Trajectory Prediction\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14956v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Autonomous driving systems rely on precise trajectory prediction to plan safe and efficient movement.\u003c/li\u003e\n\u003cli\u003eGraph Neural Networks (GNNs) have become a promising approach for modeling the spatiotemporal interactions between road agents.\u003c/li\u003e\n\u003cli\u003eHowever, designing GNN architectures for trajectory prediction remains non-standardized, with little guidance on which layers effectively capture spatial interactions and temporal dynamics.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14956v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Autonomous driving systems rely on precise trajectory prediction to plan safe and efficient movement\u003c/li\u003e\n\u003cli\u003eGraph Neural Networks (GNNs) have become a promising approach for modelling spatiotemporal interactions among road agents\u003c/li\u003e\n\u003cli\u003eHowever, designing GNN architectures for trajectory prediction remains non-standardized, with little guidance on which graph layers effectively capture spatial…\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/2606.14960\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eLeveraging Physiological Signals to Predict Exam Outcomes with Machine Learning\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14960v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: This study investigates the application of machine learning models to predict exam outcomes by utilizing physiological data collected during exams.\u003c/li\u003e\n\u003cli\u003ePhysiological stress indicators, including electrodermal activity, heart rate, and skin temperature, are analyzed to reveal their relationship with academic performance.\u003c/li\u003e\n\u003cli\u003eA variety of machine learning methods were employed, from standard models like logistic regression, random forests, and support vector machines to more advanced architectures, including transformers, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) models.\u003c/li\u003e\n\u003c/ul\u003e\n\u003c/li\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14960v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: This study investigates the application of machine learning models to predict exam outcomes using physiological data collected during examination sess…\u003c/li\u003e\n\u003cli\u003ePhysiological stress indicators, including electrodermal activity, heart rate, and skin temperature, were analyzed to uncover their association with academic pe…\u003c/li\u003e\n\u003cli\u003eA variety of machine learning approaches were employed, ranging from standard models like logistic regression, random forest, and support vector machines to mor…\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/2606.14965\"  class=\"external-link\" target=\"_blank\" rel=\"noopener\"\u003eBenchmarking Instance-Dependent Label Noise with Controlled Corruptions\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003ePublication Time: 2026-06-16 12:00 Beijing Time\u003c/li\u003e\n\u003cli\u003eAbstract: - arXiv:2606.14965v1 Announcement Type: New.\n\u003cul\u003e\n\u003cli\u003eAbstract: Synthetic instance-dependent label noise (IDN) benchmarks are widely used to evaluate noisy label learning methods, but existing methods often generate noise through imperfect annotators or classifier evaluators, thereby obscuring the source of ambiguity.\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 CILN, a benchmark generation framework that creates IDN through controlled input corruptions.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003eA diverse voter pool labels corrupted instances, producing benchmark datasets in which both the source and severity of ambiguity are explicit and controllable.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEN Key Points:\n\u003cul\u003e\n\u003cli\u003earXiv:2606.14965v1 Announce Type: new\u003c/li\u003e\n\u003cli\u003eAbstract: Synthetic instance-dependent label noise (IDN) benchmarks are widely used to evaluate noisy-label learning methods, yet existing approaches typically…\u003c/li\u003e\n\u003cli\u003eWe introduce CILN, a benchmark generation framework that creates IDN through controlled input corruptions\u003c/li\u003e\n\u003cli\u003eA diverse voter pool labels corrupted instances, producing benchmark datasets in which both the source and severity of ambiguity are explicit and controllable\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": 7284,
  "readingTime": 35,
  "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=\"#-ai-hot-topics-on-x\"\u003e🌐 AI Hot Topics on X\u003c/a\u003e\n      \u003cul\u003e\n        \u003cli\u003e\u003ca href=\"#topic-1-spacex-acquires-cursor-maker-anysphere-in-60-billion-stock-deal\"\u003eTopic 1: SpaceX Acquires Cursor Maker Anysphere in $60 Billion Stock Deal\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-2-matt-shumer-seeks-mobile-control-for-claude-code-ai-agent\"\u003eTopic 2: Matt Shumer Seeks Mobile Control for Claude Code AI Agent\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-3-openai-fixes-codex-outage-and-promises-rate-limit-resets\"\u003eTopic 3: OpenAI Fixes Codex Outage and Promises Rate Limit Resets\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-4-openai-launches-developers-plugin-for-codex-coding-app\"\u003eTopic 4: OpenAI Launches Developers Plugin for Codex Coding App\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-5-ai-builders-embrace-agentic-loops-for-self-reliant-tasks\"\u003eTopic 5: AI Builders Embrace Agentic Loops for Self-Reliant Tasks\u003c/a\u003e\u003c/li\u003e\n        \u003cli\u003e\u003ca href=\"#topic-6-anthropics-claude-managed-agents-speed-up-ai-production-deployment\"\u003eTopic 6: Anthropic\u0026rsquo;s Claude Managed Agents Speed Up AI Production Deployment\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  \u003c/ul\u003e\n\n  \u003cul\u003e\n    \u003cli\u003e\u003ca href=\"#1-todays-hot-topics-local-on-device-models-and-the-expansion-of-programming-agents-into-the-physical-world\"\u003e1. Today\u0026rsquo;s Hot Topics: Local On-Device Models and the Expansion of Programming Agents into the Physical World\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#2-unique-perspectives-and-industry-outlook\"\u003e2. Unique Perspectives and Industry Outlook\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#3-recommended-tools--resources\"\u003e3. Recommended Tools \u0026amp; Resources\u003c/a\u003e\u003c/li\u003e\n    \u003cli\u003e\u003ca href=\"#-appendix-todays-watch-list-source-update\"\u003e📚 Appendix: Today\u0026rsquo;s Watch List Source Update\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=\"#google-deepmind-blog-a_full\"\u003eGoogle DeepMind 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
}