Behavior Trees Enable Structured Programming of Language Model Agents

arXiv:2404.07439v1 Announce Type: cross
Abstract: Language models trained on internet-scale data sets have shown an impressive ability to solve problems in Natural Language Processing and Computer Vision. However, experience is showing that these models are frequently brittle in unexpected ways, and require significant scaffolding to ensure that they operate correctly in the larger systems that comprise "language-model agents." In this paper, we argue that behavior trees provide a unifying framework for combining language models with classical AI and traditional programming. …

abstract agents arxiv behavior computer computer vision cs.ai cs.cl data data sets experience however internet language language model language models language processing natural natural language natural language processing processing programming scale solve trees type vision