GA4GC: Greener Agent for Greener Code via Multi-Objective Configuration Optimization
Coding agents powered by Large Language Models (LLMs) face critical sustainability and scalability challenges in industrial deployment, often incurring costs that may exceed optimization benefits. We introduce GA4GC, the first framework to optimize coding agent runtime (greener agent) and code performance (greener code) trade-offs by discovering Pareto-optimal agent hyperparameters and prompt templates. Evaluation on the SWE-Perf benchmark demonstrates up to 135-fold hypervolume improvement, reducing agent runtime by 37.7% while improving correctness. Baseline comparisons and influence analysis confirm the effectiveness of GA4GC, identify temperature as the most influential hyperparameter, and provide actionable strategies to balance agent and code sustainability in industrial deployment.
