TokenGate: Adapting LLMs for pre-fill with learned skimming for ~11x speedup (github.com) 1 points by anandvc 3d ago ↗ HN
1 comment
[ 4.7 ms ] story [ 18.6 ms ] thread