MCTS Enhanced AI AGENTS: SELA (Stanford, UC Berkeley)

In the video “MCTS Enhanced AI AGENTS: SELA (Stanford, UC Berkeley)” by Discover AI, the presenter discusses a new AI framework called SELA that combines Large Language Model (LLM) agents with Monte Carlo Tree Search (MCTS) to optimize Automated Machine Learning (AutoML) processes. The video explains how SELA overcomes traditional AutoML limitations by utilizing a tree-structured representation of machine learning pipelines, allowing for iterative refinement and improved adaptability. Evaluations demonstrate SELA’s effectiveness in automating complex tasks, achieving a win rate of 65% to 80% against existing baselines. The video emphasizes the significance of integrating MCTS with LLM agents for enhanced performance in machine learning automation.