• saigot@lemmy.ca
    link
    fedilink
    arrow-up
    4
    ·
    20 hours ago

    If it was done with enough regularity to eb a problem, one could just put an LLM model like this in-between to preprocess the data.

    • Azzu@lemm.ee
      link
      fedilink
      arrow-up
      4
      ·
      20 hours ago

      That doesn’t work, you can’t train models on another model’s output without degrading the quality. At least not currently.

      • FooBarrington@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        7 hours ago

        No, that’s not true. All current models use output from previous models as part of their training data. You can’t solely rely on it, but that’s not strictly necessary.

      • Vashtea@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        1
        ·
        edit-2
        13 hours ago

        I don’t think he was suggesting training on another model’s output, just using ai to filter the training data before it is used.