Beliefs replace full interaction history as working context
Belief states replace the full interaction history as the agent’s working context; belief grading supervises the content of each belief state to improve performance.
Éléments favorables
Teaching LLMs to Update Beliefs for Efficient Long-Horizon Interaction
Extrait original
Beliefs replace the full interaction history as the agent’s working context, and belief grading improves performance by supervising the contents of each belief state.
Contexte
Overview of ABBEL compared to traditional recursive summarization. .