Search papers, labs, and topics across Lattice.
The paper introduces BIRDS, a framework for quantifying the biodiversity impact of LLM serving, considering both operational and embodied effects at the request level. They define a Quality-Normalized Biodiversity Impact (QNBI) metric to analyze the trade-off between ecological impact and response quality. Their analysis across various LLMs, GPUs, and regions demonstrates that biodiversity impact accumulates significantly at scale and highlights quality-aware serving optimization opportunities.
LLM serving isn't just about carbon footprints – it's quietly eroding biodiversity, and this framework finally lets you quantify the ecological cost per query.
Large language model (LLM) serving creates environmental impacts beyond carbon and water, including ecosystem damage through biodiversity-related pathways. We present BIRDS, a framework for Biodiversity Impact of Request-Driven LLM Serving. BIRDS defines request-level functional units, quantifies operational and embodied biodiversity impact, and introduces Quality-Normalized Biodiversity Impact (QNBI) to jointly analyze ecological impact and response quality. Across diverse workloads, models, GPUs, and regions, \SYSTEM{} reveals that biodiversity impact accumulates at scale and exposes actionable quality-aware serving tradeoffs.
