事情的輪廓：我們目前所處的 AI 階段與未來展望 (The Shape of the Thing)★ 85
One Useful Thing (Mollick)139 days agoOpinionWharton School professor Ethan Mollick, in his latest article "The Shape of the Thing," sketches out a clear picture of the current state of AI technological…
Notion Workers 如何利用 Vercel Sandbox 大規模安全執行未授權的第三方程式碼★ 75
Vercel Changelog139 days agoReleaseAs the Notion platform has gradually evolved into a broader ecosystem, enabling users and developers to run custom scripts and automated workflows within…
Hugging Face Hub 推出全新「Storage Buckets」儲存桶功能★ 80
Hugging Face Blog141 days agoReleaseThe official Hugging Face blog has announced the launch of "Storage Buckets" on the Hugging Face Hub. This is a cloud object storage service designed…
讓 Token 持續流動：來自 16 個開源強化學習（RL）函式庫的啟示★ 85
Hugging Face Blog141 days agoCommentaryWith the success of reasoning models such as DeepSeek-R1, reinforcement learning (RL/RLHF) has become a critical technique for improving the alignment and…
從遊戲到生物學與超越：AlphaGo 影響力的十週年回顧★ 75
Google DeepMind Blog142 days agoCommentaryIn March 2016, Google DeepMind's AlphaGo faced legendary Go player Lee Sedol in a historic match in Seoul, ultimately winning 4 to 1. The match not only…
Import AI 448：AI 研發趨勢、ByteDance 的 CUDA 寫作 Agent、衛星邊緣 AI 與 AI 戰爭的未來★ 75
Import AI (Jack Clark)142 days agoCommentaryThis issue of Import AI 448, written by Jack Clark, takes a deep dive into the latest developments in AI R&D, automated hardware optimization, and the…
Hugging Face 發布 LeRobot v0.5.0：全面擴展開源機器人學習的各個維度★ 80
Hugging Face Blog142 days agoReleaseHugging Face has officially released version 0.5.0 of its open-source robot learning library, LeRobot, under the theme "Scaling Every Dimension." Since its…
Ulysses 序列平行化：實現百萬 Token 超長上下文的模型訓練技術解析★ 78
Hugging Face Blog142 days agoTutorialAs large language models (LLMs) push the demand for long context toward the million-token scale, the VRAM of a single GPU can no longer accommodate the…
Dean Ball 談開源模型與政府控制：Anthropic 訴戰爭部案如何影響開源 AI 的未來★ 75
Interconnects (Nathan L.)145 days agoCommentaryIn this column published in Interconnects, author Nathan Lambert cites the latest observations from policy expert Dean Ball on the high-profile "Anthropic v…
AI2 推出 OLMo Hybrid：探索未來 LLM 混合架構與開源後訓練技術前沿★ 78
Interconnects (Nathan L.)146 days agoCommentaryAs large language models (LLMs) continue to evolve, the traditional pure-Transformer architecture faces physical bottlenecks in computational efficiency and…
Hugging Face 聯手 NXP：將機器人 AI 帶入嵌入式平台（資料集錄製、VLA 微調與裝置端優化）★ 75
Hugging Face Blog146 days agoTutorialHugging Face has entered into a deep collaboration with semiconductor giant NXP (NXP Semiconductors), aimed at solving the challenge of deploying advanced…
Vercel AI Gateway 正式支援 OpenAI GPT 5.4 模型★ 75
Vercel Changelog146 days agoReleaseVercel has officially announced that its AI Gateway service now supports OpenAI's latest GPT-5.4 model. Vercel AI Gateway is an LLM infrastructure tool…
Hugging Face 推出 Modular Diffusers：為擴散模型管線打造的可組合模組化積木★ 78
Hugging Face Blog146 days agoReleaseThe Hugging Face official blog has announced the launch of "Modular Diffusers" — a major architectural overhaul of its widely popular `diffusers` library. In…
PRX 第三部分：在 24 小時內訓練一個 Text-to-Image 圖像生成模型！★ 75
Hugging Face Blog148 days agoTutorialPhotoroom, the well-known AI image editing tool, recently published Part 3 of its technical blog series on Hugging Face about its in-house image generation…
Google 推出 Gemini 3.1 Flash-Lite：專為大規模智慧應用打造，史上最快且最具成本效益的 Gemini 3 模型★ 78
Google DeepMind Blog148 days agoReleaseGoogle DeepMind has officially released Gemini 3.1 Flash-Lite, the fastest and most cost-efficient version within its latest generation Gemini 3 series of…
最新開放模型彙整 (#19)：Qwen 3.5、GLM 5、MiniMax 2.5 —— 中國實驗室的最新前沿突破★ 82
Interconnects (Nathan L.)148 days agoReleaseThis is Issue #19 of the "Latest Open Artifacts" column by well-known AI industry analyst Nathan Lambert, opening with "Welcome to the year of the horse!" It…
Import AI 447：AGI 經濟學、用生成式遊戲測試 AI，以及 Agent 生態學的興起★ 75
Import AI (Jack Clark)149 days agoOpinionIn this edition of Import AI 447, Jack Clark takes readers on a deep exploration of the social and technological transformations that artificial general…
Transformer 中的混合專家模型 (MoE) 技術解析：原理、優缺點與實作挑戰★ 82
Hugging Face Blog153 days agoTutorialMixture of Experts (MoE) has become the mainstream architecture for current large language models (LLMs). This article takes an in-depth look at how MoE…
OpenEvidence 如何利用 Vercel 打造讓醫生真正信賴的醫療 AI 系統
Vercel Changelog154 days agoBusinessIn the medical field, AI "hallucinations" and uncertainty are the biggest barriers to widespread adoption. When making clinical decisions, doctors need…
知識蒸餾對中國大語言模型（LLM）到底有多重要？回應 Anthropic 的「蒸餾攻擊」觀點★ 75
Interconnects (Nathan L.)155 days agoCommentaryAnthropic recently published research on "distillation attacks," defining the practice of external developers using its API outputs to train other models as a…
代理式架構中的安全邊界 (Security boundaries in agentic architectures)★ 75
Vercel Changelog155 days agoOpinionIn the current evolution of AI applications, AI agents have advanced from simple text generation to complex systems capable of autonomous planning, calling…
Vercel 提升 Python Functions 打包大小限制至 500MB★ 75
Vercel Changelog155 days agoReleaseThe official Vercel Changelog announces that the bundle size limit for Python Vercel Functions (serverless functions) has been significantly raised to 500MB…
Import AI 446：核能 LLM、中國大型 AI 基準測試、AI 評估與政策★ 75
Import AI (Jack Clark)156 days agoCommentaryIn this edition of Import AI 446, author Jack Clark explores three highly forward-looking and interconnected topics in current AI development: Nuclear LLMs…
GGML 與 llama.cpp 正式加入 Hugging Face，攜手保障本地端 AI 的長期發展★ 95
Hugging Face Blog159 days agoBusinessA historic milestone has arrived in the open-source AI world: GGML and llama.cpp — the open-source projects founded by Georgi Gerganov that laid the…
免費訓練 AI 模型！Hugging Face 聯手 Unsloth 推出 Hugging Face Jobs 免費微調服務★ 85
Hugging Face Blog159 days agoNew ToolHugging Face's official blog has announced exciting news for the open-source AI community: Hugging Face has formed a deep partnership with Unsloth — the…
Google DeepMind 發表 Gemini 3.1 Pro：專為複雜任務打造的更智慧模型★ 85
Google DeepMind Blog160 days agoReleaseGoogle DeepMind officially released a brand-new model today (February 19, 2026): "Gemini 3.1 Pro." According to the initial official disclosure, the core…
IBM 與柏克萊加州大學推出 IT-Bench 與 MAST：診斷企業級 AI Agent 失敗原因的全新基準與框架★ 80
Hugging Face Blog161 days agoRelease### The Pain Points of Enterprise AI Agents in Production: Why Do They Keep Failing? As large language models (LLMs) have rapidly advanced, enterprises have…
代理人時代的 AI 選擇指南：不再只是聊天機器人★ 85
One Useful Thing (Mollick)161 days agoTutorialProminent scholar Ethan Mollick, in his latest article, points out that we have officially crossed beyond the era of simple "Chatbots" and entered what he…
開源模型陷入「永久追趕」：開源與閉源的差距、蒸餾、創新週期與開源的勝算★ 80
Interconnects (Nathan L.)162 days agoOpinionThis article by Nathan Lambert takes a deep dive into the tangled competitive dynamics between open-source and closed-source AI models. Lambert argues that…
Google DeepMind 於印度推動「國家 AI 合作夥伴關係」倡議，加速科學與教育發展
Google DeepMind Blog162 days agoBusinessGoogle DeepMind has announced the official launch of its "National Partnerships for AI" initiative in India, with the goal of accelerating local scientific…