[1]
C. Niu, “TokenPowerBench: Benchmarking the Power Consumption of LLM Inference”, AAAI, vol. 40, no. 38, pp. 32582–32590, Mar. 2026.