Skip to content
Live
Nov 4Wed
  1. Hugging Face Daily Papers65

    Pandora framework: Simulating personalized customer expressions in psychological therapy

    The Pandora framework generates more realistic and individualized customer expressions by modeling the evolution of psychological and relational states, reducing AI-like responses. In an evaluation of 115 cases, Pandora outperformed five baseline systems in expression similarity and disclosure consistency scores, with only 7.2% of 45 disclosures rated as AI-like expressions, significantly lower than the baseline systems' 43.3--67.2%.

  2. Hugging Face Daily Papers38

    YANAMI: A Large-Scale Multimodal Dataset and Benchmark for Multi-Round State Interaction

    YANAMI introduces a large-scale multimodal dialogue dataset and benchmark that requires models to track and implement evolving dialogue states across rounds, revealing that current unified multimodal models still struggle with this task. YANAMI-400K contains 400,000 multimodal dialogue rounds across 4 scenarios, requiring models to continuously track, store, and revise evolving multimodal dialogue states across rounds; YANAMI-Bench contains 400 multimodal rounds for system evaluation of multimodal state interaction, jointly evaluating dialogue state understanding and visual implementation.

  3. MarkTechPost

    Tencent's WeVisOne-VL unifies detection, OCR, segmentation, and Matting.

    Tencent WeChat Vision team releases open-source visual language model WeVisOne-VL, including 2B and 4B parameter versions, supporting detection, OCR, segmentation, and Matting, using a 9.6M parameter decoder to output pixel results, surpassing same-scale baselines in over 30 benchmark tests, with the 4B version achieving a 71.3 F1 score on Dense200 dense detection.

    Same story · 2 outlets

    How the accounts differ:Hugging Face Daily Papers states it uses a lightweight pixel decoder; MarkTechPost mentions a 9.6M parameter decoder.

  5. ITHome · AI65

    Meta raises $15 billion in debt financing

    Meta announced the issuance of $15 billion in preferred notes in five parts, with terms ranging from 5 to 30 years, yielding 2.2 percentage points higher than U.S. Treasury bonds; earlier reports said it would also issue euro bonds.

  7. 36Kr · AI65

    AI+Coffee Rewriting the Industrial Chain: Cost Reduction and Efficiency Enhancement and Youth Opportunities in New Consumption Scenarios

    Meike AI has reduced its store opening costs to less than one-third of the traditional model through its embodied intelligence beverage robots, and plans to establish an OPC model demonstration store in Yinzhou District; Yun Chuang Nai Che has shifted its operational mindset, using event planning to drive foot traffic, and incubating AI short video startup teams; Lu Hui Group is cultivating composite talents such as AI coffee machine operation and maintenance, and has implemented an industry-academia-research integration training system in more than 200 universities across the country.

  8. 36Kr · AI65

    SK Telecom's operating profit in the third quarter surged more than 10 times year-on-year

    SK Telecom's third-quarter operating profit surged by 1041.3% compared to the same period last year, with a net profit turning from a loss to a profit. From July to September, operating profit reached 552.3 billion won (about $4.114 billion), with revenue increasing by 13.1% to 4.49 trillion won, and net profit was 411.8 billion won (a loss of 166.7 billion won in the same period last year). AI-related new businesses helped boost performance, with AI data center sales increasing by 6.2% year-on-year, and other AI business sales growing by 27.3%.

  9. Hugging Face Daily Papers65

    MoviEdit: A framework for editing long-segment, multi-shot video

    MoviEdit framework, through cascaded bidirectional autoregressive modeling and new benchmark tests, maintains cross-scene coherence while reducing manual data annotation, significantly outperforming existing open-source models with performance comparable to industry leading models. The framework automatically generates 87K pairs of paired editing videos for cross-shot supervision, establishes a global editing draft using a lightweight bidirectional stage, and then performs autoregressive propagation and refinement.

  10. Hugging Face Daily Papers62

    Axion 1.0: An open-source foundation model for agents, simplifying implementation

    Axion 1.0 achieves competitive agent performance through a three-stage training process (two lightweight supervised fine-tuning + reinforcement learning) using fully open-source data and training pipelines. The model employs model-aware data construction and verifiable task generation, without specialized mid-training, direct preference optimization, or policy-intrinsic distillation, demonstrating excellent performance in multiple multi-agent benchmarks, comparable to leading commercial models in some tests, and openly sharing the complete development process to promote open-agent research.

  11. ITHome · AI65

    Google adjusts account settings for US K-12 schools: either enable AI search or completely disable search

    Google announces adjustments to service permissions for US K-12 schools, requiring school administrators to manually enable search functionality while also enabling AI features, or else completely disable search. Accounts for students from kindergarten to 12th grade have Google Search, the Gemini chatbot, etc. disabled by default. Schools must choose one option. Google plans to introduce a 'guided learning' mode to gradually guide students in thinking and problem-solving.

  12. Hugging Face Daily Papers

    RoboColosseum evaluates physical intelligence in large-scale real environments

    RoboColosseum evaluates general robotic strategies across different robots and environments, finding weak consistency across robot platforms, limited open-world reliability, and significant differences in zero-shot and fine-tuning performance. This distributed evaluation framework assessed five visual language action (VLA) strategies and three cutting-edge AI agents on four robot platforms (SO-101, Franka, dual-arm YAM, humanoid G1). The results show that cutting-edge agents are comparable to dedicated VLA strategies, but all systems perform poorly in open-world tasks. RoboColosseum evaluates physical intelligence in large-scale real environments.

  13. ITHome · AI38

    Anthropic engineers warn: The more accurate AI is, the easier humans lose the ability to think independently, and the more dependent they become, the harder it is to stay alert

    Anthropic Claude Code engineer Addy Osmani warns that reducing AI model output errors makes it harder for humans to stay alert, and may lead to losing independent thinking and verification ability due to reliance on AI, forming a 'cognitive surrender'. He proposes three usage levels: 'cognitive offloading', 'postponing judgment', and 'cognitive surrender', suggesting users form preliminary hypotheses before accepting AI results, actively seek rebuttal evidence, and require Agents to list autonomous decisions and hypotheses.

  14. 36Kr · AI63

    OpenAI executives say ChatGPT's search volume has reached 20% of Google's.

    The search volume on ChatGPT is approximately 20% of Google's search volume, OpenAI adds 200 million users, with 1.2 billion weekly active users, creating 1.5 billion images per week, and 250 million weekly active users in speech mode.

    Same story · 3 outlets
  15. QbitAI58

    How does openJiuwen Agent-Memory achieve audio-visual dual-stream memory, understand long videos?

    openJiuwen Agent-Memory 0.2.0 version launched multimodal memory capabilities OmniMem-Lens, which understands video events through visual and auditory dual streams, recall accuracy improved by up to 45.7%, and online Q&A latency reduced by up to 97%. This technology segments memory according to real events, establishing a three-layer memory structure to achieve offline construction and online query, demonstrating outstanding performance on the EgoLife-QA benchmark set, with longer videos showing more significant latency reduction.

  16. 36Kr · AI67

    AMD Su Tseng-Feng: Helios Rack-Scale AI System Volume Delivery in the Fourth Quarter

    AMD Chairman Su Tseng-Feng expects the deployment of Helios Rack-Scale AI Systems to significantly increase in the fourth quarter, with OEM partners like Lenovo, HPE, and Dell beginning deliveries. The first Helios cabinets for customers like Anthropic and Meta have already been put into operation, and the scale of deployment is expected to expand further in the fourth quarter.

  18. ITHome · AI65

    Yushu Wang Xingxing: 2027 Could Be the Critical Point for the Robot Industry/Models

    Yushu Technology founder Wang Xingxing stated at the 2026 World Internet Conference in Wuzhen Summit that 2027 is likely to be the critical point for the robot industry or robot models. He believes that with the release of GPT-6, physical AI might be broken through. Currently, the core challenge of embodied intelligence is the alignment between AI model input and output and real robots, and commercial service implementation will be faster than in home and industrial fields. Wang Xingxing revealed that Yushu Technology will focus on breaking through robot brain technology, has a positive view of physical AI, and seeks global cooperation.

  19. Hacker News · 300+65

    Namespace at data centers for large-scale deployment of MacBook

    Namespace deploys hundreds of MacBook Pro in data centers, achieving remote management through custom racks and IP KVM, each MacBook equipped with an external 10G Ethernet card and a cooling stand, but faces challenges such as HDMI cable breakage and FileVault password prompt disappearance, ultimately resolving them by SSH remote unlocking and recreating network services, while planning to introduce Mac Mini and Studio models of M5 and M6 to simplify management.

  23. ITHome · AI65

    Samsung Electronics plans to produce HBM-based wafers for clients such as AMD at its second semiconductor plant in Taylor, USA.

    Samsung Electronics plans to produce HBM-based wafers for clients such as AMD at its second semiconductor plant in Taylor, USA, which began construction this month and is expected to start operations in 2030. Previously, Samsung primarily supplied such wafers to its own memory business, but starting from the HBM4 generation, memory manufacturers have begun ordering them from external advanced process foundries due to performance requirements. In addition to AMD, the industry believes that the HBM-based wafers produced at this semiconductor plant may also be used for the AI ASICs created for clients by Broadcom.

    Same story · 2 outlets
  24. 36Kr · AI35

    The South Korean government and AMD continue negotiations on follow-up cooperation on AI chips

    The South Korean government and US chip giant AMD held a steering committee meeting in Seoul on Wednesday to discuss the implementation of their AI chip cooperation agreement signed earlier this year. According to the memorandum of understanding signed in July, the two sides will cooperate to establish an AI research center and heterogeneous computing infrastructure. AMD also held the 'Advancing AI World Tour' technical event in Seoul, attracting about 1300 registrations, with Korean companies such as LG Electronics participating.

  25. ITHome · AI65

    ByteDance revealed plans to widely restrict employees' internal self-funded use of Claude, Codex, and other AI tools

    ByteDance will broadly restrict employees' internal self-paid use of Claude, Codex, and other AI tools, starting this month, banning external AI access to L4 level code repositories and Lianxiu documents, covering the entire company. This move has triggered internal discussions, with Vice President of Technology Hong Dingkun responding that the consequences of leaks of L3 and above content are severe, and that promoting products should not force the disabling of other choices. The company had previously banned third-party AI tools such as Cursor and Windsurf in batches in May 2025.

    Same story · 2 outlets
  27. ITHome · AI65

    NetEase's 'Yanyun Shishisheng' to Officially Implement NVIDIA DLSS 5 on November 24

    NetEase's 'Yanyun Shishisheng' will officially implement NVIDIA DLSS 5 on November 24, becoming the first game in China to support this technology, with the North America region launching a day early. DLSS 5 is NVIDIA's new real-time neural rendering technology announced in March 2026, which takes game frames as input and uses AI models to generate realistic lighting and materials, supporting up to 4K resolution. This solution uses neural networks to real-time reconstruct complex lighting and material details, enhancing lighting and character texture while avoiding blurriness or loss of clarity.

  29. Gemini CLI releases15

    Gemini CLI v0.67.0-nightly.20261104.g89e9d25d6 released

    Gemini CLI v0.67.0-nightly.20261104.g89e9d25d6 released, fixing stream failures during extension publishing and downloading, retaining explicit gemini-2.5-flash model in dynamic model parsing, supporting SubagentStop event mapping and merging hook definitions, waiting for history replay before ACP session/load response, providing --userns=keep-id for rootless podman.

  30. Hacker News · AI hot posts65

    The revised edition of 'Generative AI Basics: A Visual Introduction to Mathematics' has been released.

    This book provides a concise derivation introduction, explaining the mathematical foundations of modern generative AI. It constructs a coherent understanding path by connecting the ideas of major generative model families (from PCA, variational autoencoders to diffusion models, GANs, etc.). The revised edition adds visual explanations, with 195 pages, suitable for researchers, practitioners, and students interested in mathematics.

  31. The Verge · AI65

    Nexus Mods requires new mods to disclose AI usage

    Nexus Mods announces that new mod uploads must declare AI usage, requiring the tag 'AI assisted' and explaining the method of use; the AI tag will be more prominent, and the platform emphasizes 'supporting human creativity' without banning AI mods, but requires transparency; authors have 60 days (until January 2, 2027) to update existing mods to comply with the requirements.

    Same story · 2 outlets
  32. Cognition Blog42

    Claude Code Security Guide: Review of November 2026 Security Incidents and Response Measures

    A series of major security incidents occurred in November 2026, including a $634 million theft from Binance, the Claude Code Cursor extension being tampered with and backdoored, and critical vulnerabilities in PHP, GitLab, and edge devices. AI accelerates attacks but cannot stop basic attack paths, and enterprises need to take immediate security measures, such as enabling anti-phishing MFA and device-bound login, auditing public attack surfaces, deploying password managers and ad blockers, and completing all device MDM and EDR management, as well as locking browser and editor permissions within 30 days.

  33. 36Kr · AI35

    New York Exchange: USD rises, AI trading revives combined with strong US employment data

    The US Dollar Spot Index rose 0.2%, with a rebound in AI trading driving stock market gains. Meanwhile, the number of US job openings reached a four-month high, while layoffs fell to a low point in over a year. The Democratic Party is widely expected to take control of the House of Representatives, with the tech industry and the AI/chip industry closely watching the election results.

  34. 36Kr · AI38

    Hot-selling keywords in the 11.11 live stream, listed companies are increasing investment in the computing power economy.

    Li Jiaqi's live stream sells 100,000 personal membership monthly cards for 0.01 yuan each, corresponding to approximately 2 trillion token call volume, AI computing power services leverage live streaming channels to directly reach a massive number of individual users. Many computing power listed companies in the A-share market are increasing their layout in the token economy track, Xingyun Technology creates token factory business, iFLYTEK's consumer-end products drive token call volume to rise steadily.

  35. 36Kr · AI50

    OpenAI hires former Uber President Gil Hesselbeck as Chief Corporate Affairs Officer.

    OpenAI fills the nearly one-year vacancy in the key global communications role, hiring Gil Hesselbeck as Chief Corporate Affairs Officer to enhance public understanding and trust in AI as it moves towards its IPO. Hesselbeck previously served as Uber's President and Chief Corporate Affairs Officer, reporting directly to Co-founder and President Greg Brockman, and will officially join in early December.

    Same story · 3 outlets