Skip to content
Live
Nov 5Thu
  1. ITHome · AI35

    Intel releases 32.0.101.9175 Game On graphics driver: supporting 'World of Warcraft: Infinite'

    Intel Graphics Software 32.0.101.9175 graphics driver released today, supporting Arc A, B series dedicated graphics and Arc integrated graphics on 1st to 3rd generation Core Ultra processors, adds support for 'Football Manager 27', 'Blade of the Shadow Zero', 'World of Warcraft: Infinite', and fixes performance and stability issues for Arc B series dedicated graphics and integrated graphics in multiple games.

  2. ITHome · AI74

    Baidu Wenxin has open-sourced the ERNIE 5.5 Flash Base base model, with a total parameter of 280B.

    Baidu Wenxin has released ERNIE-5.5-Flash-Base and opened the ERNIE 5.5 Flash pre-trained endpoint weights. This is a sparse MoE language model with a total parameter of 280B and 14B activation per token. It was pre-trained from scratch on 33.5T tokens and supports fine-tuning and inference with a context size of up to 1M tokens.

    Same story · 4 outlets

    Why it matters: The original text provides the parameters, training data size, and open-source weights of the 280B sparse MoE base, allowing readers to judge its adaptability as a fine-tuning starting point.

  4. ITHome · AI65

    Changan Automobile Co., Ltd. has launched its first painting visual AI glue application robot in China

    Changan Automobile Co., Ltd.'s first painting visual AI glue application robot has been put into operation at the Liangjiang Factory in China. The large model detection system, built based on deep convolutional neural networks, accurately locks the optimal glue application points and standardized operation trajectories, achieving a risk identification detection rate and reporting accuracy rate of over 99.5%, with an increase of about 8% in the production line JPH.

  5. ITHome · AI74

    Google's requirement for schools to enable Gemini to use search has drawn criticism from US political figures.

    Google announces that schools using its K-12 educational software must enable AI tools like Gemini, otherwise students will not be able to use Google Search and will see a message indicating that school-provided features are unavailable. Vermont Senator Bernie Sanders accuses Google of extortion, while Connecticut Senator Richard Blumenthal, former FTC Chair Lena Khan, and New York City Mayor Eric Adams also criticize the move for potential monopoly and anti-competitive practices.

    Same story · 6 outlets
  6. ByteDance Seed · Hugging Face new models58

    ByteDance-Seed/Cross-Layer-TopK-CLT released: a Cross-Layer Transcoder designed for Qwen3-8B-Base

    ByteDance-Seed released the Cross-Layer Transcoder (CLT) for Qwen3-8B-Base, which includes 36 layers × 32,768 latent variables, with 94.2B parameters, a total size of 377 GB, supporting fp32 inference, encoding by jointly scoring the top 360 activated latent variables across all layers, and reconstructing the original MLP output during decoding.

  7. QbitAI65

    spaCy author attacks MIT Astra research: modifying Qwen weights accused of being a smoke screen

    spaCy author Matthew Honnibal questions MIT Astra research, claiming its experiment modifying Qwen3.5-0.8B weights to improve math scores is misleading, mixing reading weight, writing weight, and repairing large models; Honnibal points out that Astra repairs large models by using Python programming and multiple forward tests, not directly reading weights, criticizing its presentation for blurring the boundaries of different capabilities.

  8. Hugging Face Daily Papers65

    FrontierPhysics: Evaluating Agents for End-to-End Physics Research

    FrontierPhysics is an end-to-end benchmark containing 56 expert-designed physics tasks, evaluating agents' abilities in literature review, research planning, experimentation, and paper writing, using deterministic validators and expert scoring criteria. Among 22 models and 26 configurations, no configuration passed more than 46 times out of 168 tests, with manual review showing the main weaknesses lay in insufficient rigorous reasoning and limited expert intuition, rather than physics knowledge or execution errors.

  9. Hugging Face Daily Papers65

    The TurnWise framework enhances streaming video understanding through round-level credit assignment.

    The TurnWise framework improves streaming video understanding by round-level credit assignment and decision weight balancing. Its round-level credit assignment constructs group relative advantages using prefix scores from silent completions, and round-mean policy optimizes the token loss per round. Experiments show this method outperforms existing streaming video understanding RL methods.

  10. ITHome · AI

    OpenAI acquires Pierre, expanding Codex code writing, understanding, and review tools

    OpenAI acquires the enterprise Pierre, obtaining the Code Storage storage system and the Trees, Diffs, Highlights tool libraries, to enhance Codex's code review and browsing capabilities. Pierre's Trees and Diffs tools support quick location and change review, and the functionality will be integrated into Codex and remain open source.

    Same story · 4 outlets

    How the accounts differ:36Kr states the deal targets Microsoft GitHub's code hosting business; OpenAI News and The Decoder do not mention this specific target.

  11. ITHome · AI65

    Uber executive: Currently, almost no one uses AI agents to book rides.

    Uber CEO Dara Khosrowshahi stated that during the week of Halloween in October, there were only 4 orders placed through AI agents for Uber rides, and only 500 orders for Uber Eats placed through AI agents, which accounted for a very small proportion of the platform's total orders. On average, users could complete a ride request in the Uber app in just 50 seconds, and using AI agents did not necessarily save users time.

  13. ITHome · AI

    Meta acquires AI programming startup Conductor

    Meta acquires AI programming startup Conductor, which develops tools that allow users to advance multiple software projects by deploying AI agents in multiple parallel workspaces. Conductor's cloud service will be shut down on November 20.

    Same story · 4 outlets
  14. ITHome · AI65

    Apple Vice President Ge Yue: AI will not replace apps, it will further promote the development of the app ecosystem.

    Apple Vice President Ge Yue stated at the Wuzhen Summit of the 2026 World Internet Conference that AI will not replace apps but will instead drive the next generation of apps to be more powerful and personalized. Last year, the Chinese App Store ecosystem facilitated developer operating revenue and sales reaching $562 billion, an increase of $23 billion from the previous year.

  16. MarkTechPost

    Google AI Research releases Agile Deliberation framework for gray-area image classification

    Google AI Research releases Agile Deliberation framework, which transforms fuzzy concepts into structured definitions through human-in-the-loop, for gray-area image classification. The framework is based on Gemini 2.5 Pro and Gemini 2.5 Flash API, requiring no fine-tuning, with an average F1 score 10.5 points higher than zero-shot Gemini 2.5 Flash, reaching 0.59 in paid scenarios.

    Same story · 2 outlets

    How the accounts differ:Google Research mentions user experiments reducing fatigue and anxiety; MarkTechPost focuses on F1 score improvements.

  18. Hugging Face Daily Papers65

    GameVerse: A minute-level game dataset for long-term interaction world modeling

    GameVerse includes 69.7K minute-level game clips covering 1,655 hours, 349.6 million frames, far exceeding existing datasets. Through unified scene consistency segmentation, quality filtering, camera pose estimation, and event-driven annotation, it significantly improves the reasoning duration of long-term interaction world models.

  20. ITHome · AI65

    The Wall Street Journal estimated that the storage upgrade premium for the Apple iPhone 18 Pro is about 200%.

    The Wall Street Journal reported that the storage upgrade premium for the Apple iPhone 18 Pro is about 200%. Upgrading from 256GB to 512GB, Apple pays about $66 in costs, adds $200 to the price, with a premium of about 203.03%. Upgrading to 1TB, the premium is about 212.5%. Since the launch of the iPhone and iPad, Apple has averaged over 80 cents in gross profit for every $1 in storage fees collected.

  21. Hacker News · AI hot posts50

    Users on Hacker News shared their current AI agent settings

    Users on Hacker News shared their personal paid AI subscriptions (such as Claude, ChatGPT/Codex, Cursor, etc.) and the external models they used (such as OpenRouter, local models) and coding tools (such as OpenCode, Aider). Some users used agent orchestration tools (such as Munder Difflin) or platforms (such as OpenClaw), mentioning the reasons for their choices and main application scenarios.

  22. 36Kr · AI65

    Qwen AI Office upgrades in-app editing and template capabilities, Word, Excel, PPT can be modified as needed

    Qwen AI Office upgrades in-app editing and custom template capabilities, Word, Excel, PPT can be modified as needed; users can generate files according to unit or personal format requirements, and directly modify content and style within Qwen, reducing cross-software operations and format rework; related features have been gradually opened for experience.

  23. ITHome · AI

    NVIDIA releases CUDA Toolkit 13.5: Officially supports Vera Rubin and upgrades multi-GPU communication.

    NVIDIA releases CUDA Toolkit 13.5, officially supporting the next-generation AI computing platform Vera Rubin and deploying for enterprise production environments. Updates and upgrades include GPU resource scheduling, multi-GPU communication, mathematical computing libraries, and performance analysis tools, optimizing cuBLAS for low-precision matrix operations on NVFP4 and MXFP8 to improve training and inference throughput.

    Same story · 2 outlets
  25. MIT News · AI

    MIT researchers have developed the UQ-DAAAM system, enhancing the accuracy of robotic memory.

    The UQ-DAAAM system can identify and correct errors in robotic memory descriptions, quantifying uncertainty by comparing multi-view images and optimizing object descriptions, thereby improving the accuracy of robots' perception of objects, locations, and time. The current correction success rate is approximately 92%, with an accuracy improvement of 5%. The system employs LLM to integrate multi-view descriptions and uses heuristic rules to select the best supplementary images, significantly reducing computational load without rerunning VLM.

  26. ITHome · AI35

    Sony Xperia 10 VIII/10 VII phones are receiving the Android 17 update, but without the Android iOS file transfer feature

    Sony has pushed the Android 17 update to the Xperia 10 VIII and Xperia 10 VII, but users cannot fully experience all new features. The update extends wallpaper and clock design, custom icon styles, improves the quick settings panel, and enhances screen recording functionality. The two phones will not be able to use Quick Share's cross-platform connectivity support, the upgraded AI camera assistant, and the desktop mode.

  28. 36Kr · AI

    Moonshot AI releases Kimi Hosted Agents, a hosted intelligent agent service.

    The Dark Side of the Moon releases Kimi Hosted Agents, hosting intelligent agents, providing the Harness for Agent operation as a cloud service, allowing enterprises to obtain a 7×24 hour hosted environment without needing to set up and maintain it themselves, supporting long-range task execution and multi-agent collaboration.

    Same story · 2 outlets
  29. ITHome · AI

    NetEase's "Naraka: Bladepoint" officially supports NVIDIA DLSS 5 and is the first domestic game to implement it.

    NetEase's "Naraka: Bladepoint" officially supports NVIDIA DLSS 5 and Neural Rendering technology, earlier than "Yanyun Shisixing", becoming one of the first games in the world to support this technology, while also supporting up to 6x frame interpolation. DLSS 5 enhances visual performance by using an AI model to improve texture details, skin subsurface scattering, and other screen displays, and can be used by updating the graphics driver to the latest version.

  31. ITHome · AI65

    AMD is expected to achieve its first single-quarter revenue record surpassing Intel

    AMD expects its revenue in the fourth fiscal quarter of the 2026 fiscal year to be approximately $17 billion, a year-on-year increase of 66% and a quarter-on-quarter increase of 27%, potentially surpassing Intel's estimate of $150-160 billion for the first time; the data center business accounts for 64%, with EPYC server processors and Instinct AI accelerators showing a near doubling in demand year-on-year.

    Same story · 2 outlets
  32. ITHome · AI65

    Claude Code Helps Users Bypass iFIT Subscription Wall on Running Machines

    Netizens Successfully Bypass iFIT Subscription Wall on NordicTrack Running Machine Using Claude Code, Transforming the Device's Large Screen into a Free-Use Streaming Interface. The Modification Process Took About 1.5 Days to Complete, and the Project Has Been Opened on GitHub, Integrating Apps Like Netflix, Crunchyroll. Netflix Launch Time Is Reduced to About 21 Seconds, but Only Supports Standard Definition Quality and May Affect Warranty.

  33. ITHome · AI

    Former Anthropic Claude Code product lead Cat Wu joined OpenAI as a core product management lead

    Cat Wu joined OpenAI as the Head of Core Product Management, involved in core product and agent-related work; she previously led the product planning and commercial development of Claude Code, driving the product's annual revenue to $2.5 billion and doubling its weekly active users.

    Same story · 3 outlets
  34. Hugging Face Daily Papers

    WebHarbor: Provides 153 executable website environments for GUI agents

    WebHarbor provides 153 executable, containerized web environments for stable evaluation and reinforcement learning training of GUI agents, supporting strong generalization across multi-step and post-authentication tasks. The suite includes 3,518 tasks, covering multi-step and post-authentication workflows, surpassing existing ones. The Qwen 4B-level GUI agent trained via reinforcement learning in WebHarbor demonstrates competitive performance with leading open-source GUI agents on four benchmarks, with code and data publicly available.

  35. QbitAI65

    The Taichu Yuanqi Heterogeneous Cluster Computing Power System is selected for the 2026 World Internet Conference Leading Technology Award and included in the collection 'The Charm of Science and Technology'

    The Taichu Yuanqi Heterogeneous Cluster Computing Power System is selected for the 2026 World Internet Conference Leading Technology Award and included in the collection 'The Charm of Science and Technology', listed in the key technology group for display. The system adopts a 'Master-Slave-Micro' three-level heterogeneous fusion computing power architecture, breaks through core technologies such as dynamic compromise between online computing and data space, low-precision mixed precision collaborative computing, and reduces the mainstream large model adaptation cycle to the level of days.

  36. QbitAI78

    GPT-Image-2.6 is fully launched, supporting direct GIF and SVG output

    OpenAI's GPT-Image-2.6 has been fully launched on ChatGPT, Codex, and API, available to both free and paid users, with overall generation time reduced by 50%, producing images in as fast as 6 seconds.

    Same story · 3 outlets