Unlocking Lossless speedups in LLMs via discrete diffusion, with Subham Sahoo

Unlocking Lossless speedups in LLMs via discrete diffusions Talk by Subham Sahoo, Sr. Research Scientist, MBZUAI - IFM Hosted by Saurabh Dash, Research Engineer, Cohere Labs Autoregressive models deliver high-quality generation but remain slow due to sequential decoding. Diffusion language models promise much faster inference through parallel generation, yet current systems such as Mercury2, Diffusion Gemma, and Nemotron Diffusion still lag behind frontier autoregressive models in quality. This talk presents a discrete diffusion approach that unlocks provably lossless speedups in LLM infer