Store or Recompute? Characterizing the Carbon Tradeoff of KV Cache Retention in LLM Inference

Published in 11th Workshop on Energy Efficient Machine Learning and Cognitive Computing, 2026

Recommended citation: A. Bassi, J. Wang, F. Kazhamiaka, D. S. Berger, and A. Sriraman, "Store or Recompute? Characterizing the Carbon Tradeoff of KV Cache Retention in LLM Inference," in 11th Workshop on Energy Efficient Machine Learning and Cognitive Computing (EMC2), 2026. https://openreview.net/pdf?id=SoP0rm8KVA