A Survey on Large Language Model based Autonomous Agents

The survey examines LLM-based autonomous agents that pursue objectives via self-directed instructions. It organizes existing work around four essential modules: agent profiling, memory handling, planning strategies, and action execution with tools.

Applications are catalogued in natural sciences, social sciences, and engineering, while both subjective human judgments and objective benchmarks are discussed for assessment. Updated figures illustrate planning variants and the shift from parameter learning to mechanism engineering.

Intended for researchers and practitioners, the work supplies references, classification refinements, and a living resource that tracks rapid progress in the field.