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.)154 days agoCommentaryAnthropic recently published research on "distillation attacks," defining the practice of external developers using its API outputs to train other models as a…
代理人時代的 AI 選擇指南：不再只是聊天機器人★ 85
One Useful Thing (Mollick)160 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.)161 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…
Opus 4.6、Codex 5.3 與後基準測試時代：2026 年我們該如何評估 AI 模型？★ 80
Interconnects (Nathan L.)169 days agoOpinionIn 2026, with the release of next-generation models such as Anthropic's Opus 4.6 and OpenAI's Codex 5.3, the AI community faces a fundamental challenge…
Vercel 如何為 AI 程式碼代理（Coding Agents）建構 AEO（代理引擎優化）追蹤系統★ 75
Vercel Changelog169 days agoCommentaryAs AI coding agents such as Cursor, Claude Code, and GitHub Copilot become everyday tools for developers, software vendors and SaaS companies face a new…
Vercel 評估指出：使用 AGENTS.md 定義 Agent 表現優於傳統的「技能 (Skills)」設定★ 78
Vercel Changelog182 days agoReleaseIn its latest technical blog post, Vercel shared a significant finding regarding AI Agent architecture: in their Agent Evaluations (Agent Evals), using a…
測試「Bash 是否就是你所需的一切」：評估 AI Agent 的終端機控制能力★ 75
Vercel Changelog187 days agoOpinionWith the rise of AI Agents (such as Claude Code, Cursor, etc.), the developer community has begun to rethink: do we need to design complex, dedicated APIs for…
Import AI 441：我的 AI Agent 開始工作了，你的呢？以及如何用「毒泉」污染 AI 系統★ 75
Import AI (Jack Clark)190 days agoCommentary### The Age of Practical AI Agents Has Arrived In this edition of his column, Jack Clark shares his personal breakthrough in using AI Agents. Previously, many…
Vercel 實戰經驗：為什麼我們刪除了 AI Agent 80% 的工具？★ 85
Vercel Changelog218 days agoOpinionWhen building AI applications, developers often fall into the trap of "more tools equals a smarter Agent." In early versions of Vercel's AI assistants and…
Google DeepMind 推出 FACTS 基準測試套件：系統化評估大型語言模型的真實性★ 80
Google DeepMind Blog231 days agoReleaseAs large language models (LLMs) are deployed across a wide range of industries, ensuring the "factuality" of model outputs and reducing "hallucination" has…
從 GPT-3 到 Gemini 3 的三年：從聊天機器人邁向 AI Agent 時代★ 80
One Useful Thing (Mollick)252 days agoOpinionWharton School professor Ethan Mollick, writing on his blog "One Useful Thing," offers a sharp retrospective on the most fundamental paradigm shift in AI over…
給你的 AI 一場面試：如何評估與測試 AI 的真實工作能力★ 80
One Useful Thing (Mollick)258 days agoOpinionAs AI tools (such as ChatGPT, Claude, and others) become more prevalent in the workplace, we are increasingly relying on them for decision-making advice…
2025 年末 AI 實用指南：Ethan Mollick 的主觀使用建議★ 85
One Useful Thing (Mollick)281 days agoTutorialWharton School professor Ethan Mollick has put together a highly personal and practical operating guide for the AI landscape of late 2025. He emphasizes that…
在 ChatGPT 中運行 Next.js：深入解析原生應用整合技術★ 85
Vercel Changelog286 days agoReleaseAt a time when AI-assisted development is rapidly evolving, Vercel has published a deep technical breakdown exploring how to natively integrate its React…
與魔法師共事：在參差不齊的技術前沿驗證 AI 的魔力★ 85
One Useful Thing (Mollick)319 days agoOpinionUniversity of Pennsylvania Wharton School professor Ethan Mollick, in his latest article, compares the experience of collaborating with generative AI (such as…
你可以直接用在 Transformers 的 OpenAI gpt-oss 加速妙招 🫵★ 82
Hugging Face Blog320 days agoTutorial### Background and the LLM Inference Bottleneck When running large language models (LLMs), autoregressive generation is inherently "memory-bandwidth-bound"…
大眾智能（Mass Intelligence）：從 GPT-5 到邊緣小模型，強大 AI 正在走向普及化★ 85
One Useful Thing (Mollick)333 days agoOpinionIn this article exploring "Mass Intelligence," University of Pennsylvania Wharton School professor Ethan Mollick reveals an imminent future: high-level…
TextQuests：LLM 在文字冒險遊戲中的表現究竟如何？Hugging Face 推出全新評估基準★ 75
Hugging Face Blog350 days agoReleaseHugging Face has recently introduced a new benchmark called "TextQuests," designed to evaluate the performance of large language models (LLMs) in text-based…
GPT-5：它就是能搞定一切（讓 AI 掌管任務的新時代）★ 85
One Useful Thing (Mollick)355 days agoOpinionRenowned AI scholar and Wharton School professor Ethan Mollick published a forward-looking observation about GPT-5 on his blog "One Useful Thing," titled…
GPT-5、GPT-5-mini 與 GPT-5-nano 現已支援 Vercel AI Gateway★ 95
Vercel Changelog355 days agoReleaseVercel announced in its official Changelog that OpenAI's latest generation flagship model GPT-5, along with its lightweight version GPT-5-mini and…
歡迎 GPT OSS！OpenAI 全新開源模型家族正式登陸 Hugging Face★ 95
Hugging Face Blog357 days agoReleaseThe Hugging Face official blog has announced exciting news, formally welcoming OpenAI's newly launched open-source model family — "GPT OSS." This is undeniably…
TimeScope：評估影片大型多模態模型（Video LMM）長影片理解極限的新基準★ 75
Hugging Face Blog370 days agoReleaseAs large multimodal models (LMMs) have achieved breakthroughs in image and short-video understanding, the industry has gradually shifted its attention to the…
回到未來：Hugging Face 推出 FutureBench 評估 AI Agent 的未來事件預測能力★ 75
Hugging Face Blog376 days agoRelease### What is FutureBench? As large language models (LLMs) and AI agents have rapidly advanced, traditional static benchmarks (such as MMLU and GSM8K) face a…
ScreenEnv：部署你的全端桌面 AI 代理（Desktop Agent）環境★ 82
Hugging Face Blog383 days agoNew ToolWith the rise of Anthropic's Claude 3.5 Sonnet "Computer Use" and various GUI-oriented multimodal models, "desktop agents" have become one of the hottest areas…
立即上手 AI：實用快速指南 (Ethan Mollick 著)★ 85
One Useful Thing (Mollick)400 days agoTutorialUniversity of Pennsylvania Wharton School professor Ethan Mollick recently published an extremely practical AI quick guide, "Using AI Right Now: A Quick…
在 Replicate 上運行 OpenAI 最新模型：支援 GPT-4.1、GPT-4o 與 o 系列
Replicate Blog432 days agoNew ToolReplicate, the well-known AI model hosting and deployment platform, has announced a major update: it now officially supports OpenAI's latest-generation models…
性格與說服力：從 AI 的「迎合效應」中學習★ 75
One Useful Thing (Mollick)453 days agoOpinionWharton School professor Ethan Mollick, in his latest article "Personality and Persuasion," delves into AI's persuasive power and the psychological mechanisms…
OpenAI 發表 o3、o4-mini 推理模型與開源終端機工具 Codex CLI★ 90
TLDR AI (Buttondown)467 days agoReleaseOpenAI recently held a live stream and published a blog post to officially announce the new reasoning model o3 and the lightweight reasoning model o4-mini…
介紹 HELMET：全面評估長文本語言模型（Long-context LLMs）的新一代基準測試★ 80
Hugging Face Blog468 days agoRelease### Background and Pain Points: Moving Beyond the Overly Simple "Needle in a Haystack" Test In recent years, the context window length supported by large…
OpenAI 推出全新主力模型 GPT 4.1：效能與實用性的新平衡★ 85
TLDR AI (Buttondown)469 days agoReleaseOpenAI has officially released its new flagship model GPT 4.1, positioned as the next-generation "workhorse" designed to give developers and enterprises the…