• intensely_human@lemm.ee
    link
    fedilink
    English
    arrow-up
    4
    ·
    1 year ago

    Someone once posted that they got GPT to produce much higher quality output by instructing it to make an argument, critique that argument, make a new argument with the critique in mind, and so on.

    Maybe stable, precise reasoning is just a result of multiple language models arguing with each other.

    • G234323@lemmy.fmhy.ml
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      yeah probably asking for more detail helps. GPT is limited in output length, so arguing and going to the detail will produce better output