skunnavakkam comments on Redundant Attention Heads in Large Language Models For In Context Learning

[skunnavakkam]

norm \in \mathbf{R}, doesn’t matter