Engineering Sustainable Agents: A Systematic Comparison of Agentic LLMs for Developer Workflows
What it is
This paper presents an empirical study comparing non-agentic and multi-agent LLM configurations across five software engineering tasks. It evaluates accuracy, latency, and energy consumption to highlight the trade-offs between agentic complexity and sustainability.
The problem it solves
Agentic LLM systems often incur significantly higher computational and environmental costs without proportional accuracy gains.
What you could use it for
- Designing energy-efficient LLM-based development tools
- Selecting appropriate agent architectures for specific coding tasks
- Benchmarking the environmental cost of multi-agent workflows
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.