2026-11-17 AI 日报
统计日时区:America/Los_Angeles · 尚未结算
google-research/paper-orchestra · Tencent/LoopForge · elementalsouls/Claude-BugHunter
开源项目
- google-research/paper-orchestra
PaperOrchestra is a multi-agent framework for automated AI research paper writing.
- Tencent/LoopForge
Self-hosted workflows for coding-agent delivery. Orchestrate CodeBuddy, Codex and Claude Code with scripts and human approval gates. Includes Local workflows…
- elementalsouls/Claude-BugHunter
A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core…
- jingyaogong/minimind
🧠 Train a 64M-parameter LLM from scratch in just 2h!
- dream-num/univer
Build your own AI office with Univer Office SDK. Bring editing, calculation, collaboration, and file processing into your product for people and AI agents.…
- mvanhorn/last30days-skill
AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary
- affaan-m/ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode,…
- stablyai/orca
Orca lets you run a fleet of coding agents (e.g. Claude Code, Codex) with your own subscription. Available on desktop, mobile and remote runtime.
- alibaba/open-code-review
Secure, fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level…
- google-labs-code/stitch-skills
A library of Agent Skills designed to work with the Stitch MCP server. Each skill follows the Agent Skills open standard, for compatibility with coding agents…
论文
- Sage S2.1 Flash Technical Report
We introduce Sage S1.5 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over scientific evidence and act in…
- InclusionMed: Benchmarking AI for Inclusive Healthcare
Advancing medical AI for inclusive healthcare requires evaluation across diverse healthcare roles, tasks, and care settings, but benchmark coverage remains…
- AllocFlow: End-to-End Portfolio Trading Action Generation from Raw Market Data
Real-world portfolio management requires converting noisy market observations into coordinated, simultaneous allocations across multiple assets under shared…
- SWE-CoWork: Benchmarking Coding Agents as Digital Employees in Organizations
A coding agent working within a software organization must establish what to build, collaborate with colleagues, and verify what it delivers. We introduce…
- SPILLWAY: Separating Target Admission from Sanitization to Defend Tool-Using Agents against Prompt Injection
Tool-using agents act on email, files, and payments, so an instruction hidden in a web page or a shared document can make an agent send data or money to an…
- VoxMel TTS Technical Report
Continuous autoregressive TTS relies on an acoustic target that enables stable next-latent modeling while preserving fine-grained acoustic details. While…
- K2-Horizon: A Frontier Fleet of Open Source Language Models
We present K2-Horizon, a fleet of six fully open language models, from a 0.9B dense model to a 375B-A23B MoE, built for general-purpose reasoning and agentic…
- MovieRoll: A Systematic Framework for Rolling Out Long-Form Videos
Streaming video generation must sustain quality beyond the training horizon while handling self-generated history and growing cache costs. To address these…
- ScaleVLA: Scaling Vision-Language-Action Models to 552 Billion Parameters
Frontier Vision-Language Models (VLMs) such as GPT-6 Astra have demonstrated remarkable capabilities in robot control, suggesting that broad pretraining…
- LynnReal-Hand: Unlocking World Interaction with Hand Control
Interactive video world models enable exploration through text and camera controls, yet direct object interaction requires finer control. Hands provide a…
行业新闻
- GPT‑6.2 Astra and Sol
- How the Claude Code team thinks about verification
- Harvard researchers find AI coding agents shifted the bottleneck to code review
- Not Everything Needs an LLM
- Advancing computer-use agents for healthcare teams
- Regulatory Capture by Anthropic Through METR
- Why is Gemini 4 Carbon taking so long?
- Effects of a Nuclear Detonation in Low Earth Orbit
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that…
- Show HN: TrnRT – An open-source LLM inference engine for AWS Trainium
Hi HN, Sanchit here from RunAnywhere (YC W26). We just open-sourced TrnRT: Trainium RunTime, our inference engine for AWS Trainium. We started 3 weeks back,…
- GPT-6.2 Astra scores 100% on ARC-AGI-3
社区动态
- To celebrate Mistral Vibe + Le Chonk becoming the world’s best open model-harness duo, we’re making Vibe free in…
To celebrate Mistral Vibe + Le Chonk becoming the world’s best open model-harness duo, we’re making Vibe free in Europe for the first 3 months. (Yes, it’s…
- GPT-6.2 Astra from @OpenAI on ARC-AGI (Verified): - ARC-AGI-3: 100.0% (SOTA), $4.1K (provider adapter), 70.3%, $24.2K…
GPT-6.2 Astra from @OpenAI on ARC-AGI (Verified): - ARC-AGI-3: 100.0% (SOTA), $4.1K (provider adapter), 70.3%, $24.2K (standard) - ARC-AGI-2: 95.8% (SOTA),…
- With Carbon's help, I ran Gemma directly in an office printer 😅🖨️ 🤏Carbon compiled a small transformer engine +…
With Carbon's help, I ran Gemma directly in an office printer 😅🖨️ 🤏Carbon compiled a small transformer engine + quantized Gemma 3 270m into a single…
- "...Claude figured out how to train a 6M parameter model that fits into 3.2 MB optimized for Game Boy. The result: on-cartridge generative conversation with a finch oddly knowledgeable about the Revolutionary War", Tim Hwang
- Intel Shares First Benchmarks of Its Next-Gen Crescent Island "D270" GPU, Claiming a 5.9x Throughput Over RTX PRO 6000 As Xe3P Goes Big On FP64
The title talks about FP64. For local inference this info from the content is relevant: "In AI benchmarks, Intel's D270 GPU offered 29% lower latency and…
- ICLR 2027: PCs ran an AI detector on reviews, peer review is cooked [D]
The PCs ran AI detectors on the reviews and left this comment on undisclosed AI usage in reviews: AI use in this review An AI-text detector assesses this…
- OVH announces Olala mini 7A1B
Olala mini is based on a 7-billion-parameter architecture (7B), with one billion active parameters spread across 256 experts per layer, and hybrid attention…
- nvidia/Qwen3.8-Flash-Next-NVFP4-QAD
Looks pretty interesting! Also sorry if the "new model" flair is inappropiate here, wasn't sure if I should pick that or "discussion".
- For those running LLMs in banks or other regulated companies: what did your security/compliance team require before approving it?
Disclosure up front: I work on a self-hosted LLM gateway for exactly this use case, so I'm biased. Not linking it here, I'm more interested in how others…
- Concerns About Peer Reviewers Inflating OA Scores [D]
The paper is largely incremental, with very little improvement over existing work. One of the reviewers gave a one-line weakness that isn't even a real…
官方发布
- Introducing GPT-6.2 Astra and Sol
Introducing GPT‑6.2 Astra and Sol, with better professional outputs, coding, computer use, and efficiency at lower cost across ChatGPT Work, Codex, and the API.
- Gemini 4 Hydrogen: a faster model for general purpose robotics
Google's Gemini 4 Hydrogen brings strong spatial and embodied reasoning capabilities to a fast, efficient model built specifically for robotics. To help…
- codex 0.170.0
New Features Added WebRTC negotiation for realtime voice calls in the TUI. (#61172) Added automatically refreshed authentication headers for OpenTelemetry…
- Voice editing in Gemini 3.5 Transcribe
gemini-3.5-transcribe and gemini-3.5-transcribe-live now support voice-driven editing, rewriting, and formatting of dictated speech or pre-existing text…
- Find Ways to Save Time and Money with Tasks in Gemini, now available free of charge
What: Starting now, Tasks in Gemini is rolling out free of charge for hundreds of millions of users in the US, India, Canada, Australia, Japan, and Korea,…
- We’ve launched Linked billing in beta with the user-profiles-billing-2026-10-01 beta header.
We’ve launched Linked billing in beta with the user-profiles-billing-2026-10-01 beta header. Your customers can now pay for the Claude usage in your app…
- Choose a release channel in Claude Desktop
You can now use a release channel to set which version of Claude Desktop you get when the app updates. Stable stays the default and gets new versions less…
- Introducing Grok Bot 1.0
- Advancing computer-use agents for healthcare teams
- Muse Is Now Available in Australia: The World’s First Personal AI Agent Built for Everyone
Muse is available in Australia from today, with local connector experiences across creativity, shopping, property, fitness and entertainment rolling out…