The field of large language models (LLMs) is rapidly advancing, with a focus on integrating these models with domain-specific knowledge to improve planning and problem-solving capabilities. Recent developments have explored the use of LLMs in assisting classical planners, enhancing tool planning, and generating immersive visualizations. Notably, researchers have proposed novel paradigms for utilizing LLMs, such as integrating domain-specific knowledge to ensure valid plans, and constructing request-specific tool graphs to select tools efficiently. Additionally, LLMs have been used as visualization agents for immersive binary reverse engineering, and to improve generalized planning through strategy refinement and reflection.

Some noteworthy papers in this area include: Inspire or Predict? Exploring New Paradigms in Assisting Classical Planners with Large Language Models, which proposes a novel LLM-assisted planner integrated with problem decomposition. GTool: Graph Enhanced Tool Planning with Large Language Model, which constructs a request-specific tool graph to select tools efficiently and generates sufficient dependency information understandable by LLMs. Mapping the Course for Prompt-based Structured Prediction, which combines LLMs with combinatorial inference to address hallucinations and complex reasoning problems. Self-Guided Function Calling in Large Language Models via Stepwise Experience Recall, which proposes a self-guided method for function calling in LLMs via stepwise experience recall. DeepThink3D: Enhancing Large Language Models with Programmatic Reasoning in Complex 3D Situated Reasoning Tasks, which enhances the ability of LLMs to perform complex reasoning in 3D scenes.