ARTFEED — Contemporary Art Intelligence

AutoRPA: LLM-Driven GUI Automation Framework

ai-technology · 2026-05-22

AutoRPA is a framework that combines Large Language Model (LLM) agents with traditional Robotic Process Automation (RPA) to automate repetitive GUI tasks efficiently. It addresses the inefficiency of ReAct-style agents that repeatedly invoke LLM reasoning for each step. AutoRPA distills decision logic from ReAct agents into robust RPA functions using two core innovations: a translator-builder pipeline that converts hard-coded actions into soft-coded procedures, and a hybrid repair strategy for robustness. The framework aims to reduce manual effort in RPA development while maintaining runtime efficiency.

Key facts

  • AutoRPA bridges LLM agents and traditional RPA.
  • It distills ReAct-style agent logic into RPA functions.
  • Uses a translator-builder pipeline for code synthesis.
  • Employs a hybrid repair strategy for robustness.
  • Aims to reduce manual effort in RPA development.
  • Focuses on repetitive GUI tasks.
  • Improves runtime efficiency over ReAct paradigm.
  • Proposed in arXiv paper 2605.21082.

Entities

Institutions

  • arXiv

Sources