New prompting method uses thought templates | Buffer of Thoughts

This new paper proposes a new prompting method that leverages thought templates to enhance the accuracy and efficiency of LLM-based reasoning.
Paper: arxiv.org/abs/...
#ai #machinelearning #llms #promptengineering

[@rupjitchakraborty8012]

Thank you so much for taking out time to explain this

[@rupjitchakraborty8012]

Thanks!

[@IdPreferNot1]

All these developments seem to just come from the stateless nature of LLMs. Using the context window to load data context, reasoning approaches, etc...
With the advent of million+ token windows, have you seen anything beyond this formalizing something like a "LLM boot package"?