AI Infrastructure SoftwareDhiraj Kalamkar et al. (Intel)
Intel presents A Study of BFLOAT16 for Deep Learning Training
Presented A Study of BFLOAT16 for Deep Learning Training — the first comprehensive empirical demonstration that BF16 training is effective
Why it matters
A format that removed FP16's scaling burden. Later the de facto default training precision for the TPU, Ampere and Hopper generations
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/abs/1905.12322Added in Week 35, 2026 · See that week