• henrique@lemmy.world
    link
    fedilink
    arrow-up
    10
    ·
    1 year ago

    Made an integration with one LLM recently and had the same thought. Too many inputs and the responses are incorrect. Too few and it starts to hallucinate.