From Imperative to Declarative: Towards LLM-friendly OS Interfaces for Boosted Computer-Use Agents

📰 ArXiv cs.AI

Researchers propose declarative OS interfaces to improve large language model-powered computer-use agents

advanced Published 26 Mar 2026
Action Steps
  1. Identify the limitations of imperative GUIs for LLM-powered agents
  2. Design declarative OS interfaces that allow LLMs to specify high-level goals directly
  3. Implement and test the declarative interfaces to measure their impact on agent performance
  4. Refine the interfaces based on experimental results to optimize LLM calls and success rates
Who Needs to Know This

AI engineers and researchers working on LLM-powered agents can benefit from this proposal to improve the efficiency and success rate of their agents, while software engineers can apply these concepts to design more LLM-friendly OS interfaces

Key Insight

💡 Declarative OS interfaces can simplify LLM interactions with computers by allowing high-level goal specification, reducing errors and LLM calls

Share This
🤖 Declarative OS interfaces can boost LLM-powered computer-use agents! 💻
Read full paper → ← Back to News