hhuoobb
V2EX  ›  随想

Reflection 70B 号称世界上最好的开源大预言模型

  •  
  •   hhuoobb · Sep 7, 2024 · 2503 views
    This topic created in 638 days ago, the information mentioned may be changed or developed.
    benchmark 里面, reflection 70b 各个方面的能力都比 Claude 3.5 Sonnet, Claude 3 Opus , GPT-4o Gemini 1.5 Pro , Llama 3.1 405B 要强。
    我用了下感觉这个模型好像没有这么牛啊。不知道是不是使用姿势不对。
    比如我发 helo, 它会长篇大论的解释“helo"是一个西班牙语打招呼的词,看起来很学术的样子。
    但是最后一句居然带着提示词。这开发者不知道是怎么搞的把提示词放在回答里了。

    相关链接:
    https://huggingface.co/spaces/featherless-ai/try-this-model
    https://reflection70b.com
    7 replies    2024-09-08 17:18:43 +08:00
    tool2dx
        1
    tool2dx  
       Sep 7, 2024 via Android
    这种在 AI 里叫过拟合,光跑分强没用啊,还是要实战强才行。
    Chihiro316
        2
    Chihiro316  
       Sep 7, 2024
    大预言模型。。。可以预言我能活多久么[doge]
    line
        3
    line  
       Sep 7, 2024 via iPhone
    把 cot 硬微调到 模型里
    jlak
        4
    jlak  
       Sep 7, 2024 via iPhone
    跑分考题就那些,全喂给 AI 再训练出不得满分💯
    Sting1226
        5
    Sting1226  
       Sep 7, 2024
    能语言我什么时候财务自由?
    PositionZero
        6
    PositionZero  
       Sep 7, 2024
    随便让它解释个名词,它一边回复一边道歉:

    ```
    ... Actually, that's not entirely accurate. A shadowban doesn't make content completely invisible, but rather reduces its visibility significantly.

    ... I made a mistake here. While this is true for some platforms, it's not universal. The effects of a shadowban can vary between platforms, and I shouldn't generalize.
    ```
    hhuoobb
        7
    hhuoobb  
    OP
       Sep 8, 2024
    @Chihiro316 新换了个输入法 打错了[doge]
    About   ·   Help   ·   Advertise   ·   Blog   ·   API   ·   FAQ   ·   Solana   ·   3085 Online   Highest 6679   ·     Select Language
    创意工作者们的社区
    World is powered by solitude
    VERSION: 3.9.8.5 · 40ms · UTC 13:47 · PVG 21:47 · LAX 06:47 · JFK 09:47
    ♥ Do have faith in what you're doing.