Researcher, Intel
1 paper at NeurIPS 2025
We demonstrate for the first time, fully quantized training of a 7B LLM using FP4 format.
