DiTCarbon: Predictive Carbon Footprint Estimation for Diffusion Transformer Inference
Trustworthy AI4Good Workshop, ICML 2026
Predicts the energy use and carbon emissions of AI image and video generation. Cuts energy-prediction error by 36% versus a runtime-informed baseline and 59% versus a learned baseline.
@inproceedings{maravilla2026ditcarbon,
title = {DiTCarbon: Predictive Carbon Footprint Estimation
for Diffusion Transformer Inference},
author = {Maravilla, Jon and Kur, Justin and Zhao, Kaiqi},
booktitle = {ICML 2026 Workshop on Trustworthy AI4Good},
year = {2026}
}