If a task isn't done well and I keep asking the model to revise it in the current conversation, it never gets better. But starting a new conversation and having an untouched model step in, it gets solved in no time.

Maybe this is what they call 【being too close to the problem to see clearly】.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

2 Likes