Serving Masked Diffusion LLMs: Characterization and Design Principles from Real Hardware
Published in Proceedings of the Workshop on AI on HPC: Performance Engineering, Challenges and Opportunities (AI on HPC), held with SC26, 2026
Recommended citation: Amin, F., Afroz, S., Moghadampanah, M., & Nikolopoulos, D. S. (2026). "Serving Masked Diffusion LLMs: Characterization and Design Principles from Real Hardware." In Proceedings of the Workshop on AI on HPC: Performance Engineering, Challenges and Opportunities (AI on HPC), held in conjunction with SC26, Chicago, IL.
Download Paper | Download Slides | Download Bibtex