Tag
The paper introduces SkillConsist, a method using bidirectional graph alignment to detect inconsistencies between declared and implemented behavior in LLM agent skills, achieving strong F1 improvements over baselines.
Object Aligner is an open-source Python library that deterministically scores two JSON objects by recursively aligning their trees, using Hungarian algorithm for unordered collections and sequence alignment for ordered ones. It introduces referential alignment for graphs/hypergraphs and can be used as a reward function in LLM prompt optimization.