跳到正文
6月13日周六
6月12日周五
  1. Rohan Paul56

    atomic[.]chat 在单张 H100(FP8)上对比 DiffusionGemma 26B A4B 与 Gemma4 26B A4B 在事实性写作任务中的表现。DiffusionGemma 速度达 763 tok/s(3.7 秒),是 Gemma4(218 tok/s,15.1 秒)的 4 倍,但错误率显著更高。在 Steve Jobs 传记、Tetris 历史和 BeOS 故事三项任务中,Gemma4 答对 45 个事实、错 5 个;DiffusionGemma 仅对 33 个、错 28 个。主题越冷门错误越多:Jobs 4 错、Tetris 12 错、BeOS 12 错,例如将 Jobs 母亲写为 Clara Clley、为 Tetris 发明者虚构同事 Geri Gulovik、将 BeBox 价格误报为 $9,999(实价 $1,600)。原因在于 DiffusionGemma 一次生成 256 tokens 并多轮抛光,只追求文本流畅性而非事实准确性。Google 官方也建议在事实重要时使用常规 Gemma4。

    引用atomic.chat@atomic_chat_hq

    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.

6月11日周四
  1. SiliconFlow71

    @NousResearch 发布了 Hermes Agent Desktop——现在通过 @硅基流动 SiliconFlow 使用前沿开源模型更加便捷 🔥 → 一键随时切换模型——DeepSeek-V4、GLM-5.1、Kimi-K2.6、MiniMax-M3 等,均在硅基流动上 ... ... 通过硅基流动开启你的 Hermes 之旅的完整指南 👇🧵

    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。

  2. jason71

    开发者 @trq212 展示如何用 Fable 通过代码和工具调用完成其自身启动视频的编辑:调用转录服务、ffmpeg、调色、Figma MCP,制作 Remotion UI 并渲染,全程未手动操作视频编辑器。主推文作者 Jason Liu 幽默总结了一个“向 Anthropic 设计团队请求最佳实践”的流程,但对 Fable 的用法表示赞赏。

    引用Thariq@trq212

    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.

6月10日周三
  1. meng shao69

    Wix VP @IShmool 推出的免费课程「Zero to Claude Code」现已支持简体中文,作者专门用中文官宣并感谢中文 AI 社区。该课程面向零基础学习者,讲解如何用 Claude Code 从零构建真实产品。已有 300 名中国学生参与学习。课程完全免费,访问 zero2claude.dev 即可开始。

    引用itay shmool 🇮🇱☮️✨️@IShmool

    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。

  2. Nathan Lambert63

    Claude Fable 5 在 APEX-SWE 软件工程评测中取得 65.5% Pass@1 总体成绩,较 Claude Opus 4.8 高约 18 个百分点。两个子类别中,Integration 为 61.3%,Observability 高达 69.7%,后者比 Opus 4.8 领先 26 个百分点。Fable 5 是首个在 Observability 类别突破 50% 的模型,也是唯一在该项上得分高于 Integration 的模型(其他模型均相反)。Observability 此前一直是所有模型的瓶颈,Fable 5 首次打破这一局面。主推文认为,虽然模型 token 价格不菲,但对大量企业而言物有所值。

    引用Mercor@mercor_ai

    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.

6月9日周二
  1. Tibo66

    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?

    引用Peter Steinberger 🦞@steipete

    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。