策略梯度推导: https://rlhfbook.com/c/06-policy-gradients#deriving-the-policy-gradient
derivation of Policy Gradient.
策略梯度推导: https://rlhfbook.com/c/06-policy-gradients#deriving-the-policy-gradient
derivation of Policy Gradient.
导出 YouTube Studio 元数据的最佳方式是什么?@OfficialLoganK 作为创作者,你们有 WebMCP 或 YouTube 连接器的计划吗?
Diffusion Gemma is 4x faster, but makes 6x more mistakes! We benchmarked the new diffusion LLM against its autoregressive twin on a single H100 (FP8). We gave each the same three tasks: write a Steve Jobs biography, the history of Tetris, and the story of BeOS - every next topic less popular than the previous one. Then we fact-checked every claim in every answer. Gemma4 got 45 facts right, 5 wrong. DiffusionGemma got 33 right, 28 wrong. The less popular the topic, the worse it got: 4 mistakes on Jobs, 12 on Tetris, 12 on BeOS. It named Clara Clley as Steve Jobs' mother, invented a colleague for Pajitnov named Geri Gulovik and priced the BeBox at $9,999. The real one cost $1,600. Outputs: Gemma4 26B A4B: 218 tok/s · 15.1s total · 45 facts · 5 mistakes DiffusionGemma 26B A4B: 763 tok/s · 3.7s total · 33 facts · 28 mistakes The reason is simple. DiffusionGemma throws 256 tokens on the screen at once and polishes them pass after pass until the text sounds smooth. Smooth is all it cares about: a fake name, date or number sounds just as smooth as a real one, so it stays. Regular Gemma4 meanwhile writes one word at a time and checks every new word against everything before it. Google says it themselves in the launch post: quality is lower, use regular Gemma 4 when facts matter.
http://x.com/i/article/2065098287758348288
http://x.com/i/article/2064864205749280768
http://x.com/i/article/2065151123128721408
推荐理由:虽然 Hermes Agent Desktop 不是新物种,但 SiliconFlow 这趟手把手教程让切换国产开源模型变得极其顺手,想做桌面 Agent 的读者可以立刻跑起来。
@jakedahn I've been writing a whole guide! https://simonwillison.net/guides/agentic-engineering-patterns/
Lots of people asked how I used Fable to edit its own launch video so I made a video about that! TLDR it wrote a lot of code & tool calls to use transcription services, ffmpeg, do colorgrading, use the figma mcp, make remotion UI and render it. I didn't touch a video editor.
http://x.com/i/article/2064857003743391744
Lots of people asked how I used Fable to edit its own launch video so I made a video about that! TLDR it wrote a lot of code & tool calls to use transcription services, ffmpeg, do colorgrading, use the figma mcp, make remotion UI and render it. I didn't touch a video editor.
Go #MessiMode 上传一张你的照片并尝试这个提示词:“将我的头发变成本国国旗的颜色,但要看起来自然。如果没有提供国家或图片,请询问。”
推荐理由:一个官方给出的趣味 prompt,让你把头发染成国旗色,虽然不是什么大更新,但胜在简单好玩,普通人也能立刻上手玩一下。
有人用 Claude Fable 5,基本上通过一个提示词,在 Three.js 中构建了这个漂亮的类《孤岛危机》游戏。 Credit: @dangreenheck
一则巧妙的角色反转短片,凸显了AI视频的创意力量。幽默、视觉锐利、执行完美。干得漂亮 @ai_am_furufuru !
「The Fish Strikes Back/魚の逆襲」 #seedance #pixversecpp @PixVerse_
Introducing text-to-lottie: an open source skill and harness for generating production ready Lottie animations with codex/claude code. $ npx skills add diffusionstudio/lottie Prompts guide and repo in the comments.
zero2claude 正式推出简体中文版。 已有 300 名中国学生在学习用 Claude Code 从零开始构建真正的产品。 感谢 @shao__meng 和中国 AI 社区的支持 🇨🇳 免费课程,零基础也能上手。
“我们过去会检查 Claude 是否把工作做对,比如复核它的输出、发现它提前停止等等。而有了 Claude Fable 5,我转而检查 Claude 是否在做对的工作” - Thariq (@trq212) Claude Code
推荐理由:Claude Code团队的实战建议,把Claude从“执行工具”升级为“思考伙伴”,用/goal和Workflows实现自我验证,这套工作流比新功能本身更有价值。
Claude Fable 5 takes #1 on APEX-SWE: 65.5% Pass@1 overall. It scores ~18pp higher than Opus 4.8. We tested @claudeai Fable 5 on APEX-SWE which measures whether AI models can do real software engineering work. Fable 5 tops our two APEX-SWE categories: - Integration: 61.3% - Observability: 69.7% The standout is Observability at 69.7%, 26pp ahead of Claude Opus 4.8. It is the first model to clear 50% on the category, and the only one that scores higher on Observability than on Integration. Every other model shows the reverse. Observability has been the bottleneck for every model we have measured. Fable 5 is the first to break it. Congrats to the @AnthropicAI team.
Claude Fable 5 改变了我们 Claude Code 团队的日常运作方式。 我们过去常常验证 Claude 是否正确完成了工作。现在我们验证它是否在做正确的工作。 以下是最大的三个变化:
推荐理由:用 Cursor 又想用 OpenRouter 上 Claude 4.6 或 Llama 4 的开发者,这篇指南帮你打通关键配置,不是大新闻但值得收藏
23,000+ 篇 ChinaRxiv 论文现已免费提供,并带有更完整的英文翻译,源于一位开发者用 GPT-5.5 替换了复杂的 OCR 管道。
http://x.com/i/article/2059815427484655622
推荐理由:UniRL把扩散和LLM的强化学习塞进同一个训练循环,外加两个新算法,多模态对齐的研究者可以立刻fork代码试起来。
Since my last post, I reduced ty’s retained memory by another 15% with Codex. We're now at a ~25% memory reduction overall via /goal, largely in the background. I love working with the GPT models so much.
每月提醒:你不应再手动提示编码智能体了,而应设计循环来驱动它们。有人已经在写嵌套循环了吗?
这是你每月的提醒:你不应该再去给编码代理写提示词了。 你应该设计循环,让循环去给代理写提示词。
推荐理由:这推文把AI编程的使用方法往前推了一步,从'写好一句prompt'变成'设计一套循环'。如果你还在手动给Cursor发指令,这条该看一下。
/teach is live Learn anything, from rubik's cube to vocal harmonies to software fundamentals. npx skills add mattpocock/skills --skill teach Best skill I've ever built, video coming soon https://github.com/mattpocock/skills/tree/main/skills/productivity/teach
以下是在 Cursor 中运行的 3 个循环 1. 不稳定测试清除器 /loop 运行测试套件 20 次,收集所有间歇性失败,修复或隔离不稳定测试,直到连续 5 次完全通过才停止。
The Product Design Codex 插件与 Figma 配合最佳!可即时对设计进行迭代,然后将其导出到你的 Figma 画布。
Turn code into designs your whole team can edit by connecting Codex to the Figma canvas