Run the same experiment using four LLMs

  • Why four LLMs?
    • Enough for a supermajority vote
    • Also enough to cause a tie
    • Smallest possible number
  • Save all the LLM responses into individual files
    • one per LLM per experiment
    • keep this as a separate task
    • allows you do any kind of analysis you want in the future