This is Chapter 7 of 'The Learn Arc' series, delving into Active Inference, POMDPs, sophisticated planning, and Dirichlet learning. It significantly advances the understanding of hierarchical agents and belief-propagated tree search, making it a pivotal part of the series.
PIVOT (Plan-Inspect-eVOlve Trajectories) addresses plan-execution misalignment in LLM agents using a self-supervised framework. It iteratively refines trajectories through environment interaction, demonstrating state-of-the-art performance in empirical evaluations.
This research investigates using an agentic language model feedback framework to generate high-quality planning domains from augmented natural language descriptions. It evaluates the impact of various symbolic feedback mechanisms, like landmarks and plan validation output, in conjunction with heuristic search over model space to optimize domain quality.
This work investigates how to encode factored tasks, a classical planning representation, into SAT. It proposes several strategies for translating the factored transition relation into propositional logic and analyzes the impact of task transformations on SAT-based planner performance.