Cognitive Models and AI Algorithms Provide Templates for Designing Language Agents

Researchers propose using cognitive models and AI algorithms as templates for designing modular language agents that combine multiple large language models. The position paper formalizes agent templates that specify roles for individual LLMs and how their functionalities should be composed to solve complex problems beyond single model capabilities.

• →Contemporary large language models still struggle with difficult problems that require combining multiple LLMs into modular systems.
• →Cognitive models and AI algorithms can serve as blueprints for designing effective language agent architectures.
• →Agent templates that specify LLM roles and composition methods offer a structured approach to multi-model systems.
• →Existing language agents in literature already demonstrate templates derived from cognitive science and AI algorithms.
• →This approach aims to create more effective and interpretable language agents through established cognitive frameworks.