Tagged: attention-mechanisms

2 articles

Latest

Longformer Cuts Transformer Memory from Quadratic to Linear for Long Docs

Microsoft Drops DIFF Transformer V2, Fixing the 100x Instability That Sank V1