Understanding Thought Structures
Maciej explores the intricacies of the "tree of thought" concept, emphasizing the varying definitions of a "thought" across different designs. He clarifies that a thought can represent anything from a single step of reasoning to an entire prompt, highlighting the importance of granularity in understanding outputs from large language models. The discussion also touches on how different frameworks allow users to parameterize and control these thought processes effectively.In this clip
From this podcast
Data Skeptic
Graphs for HPC and LLMs
Related Questions