Attention Drift Collection Models trained as a part of the "Attention Drift: What Speculative Decoding Models Learn" paper, shared for reproducing experiments. • 14 items • Updated about 8 hours ago
Attention Drift: What Autoregressive Speculative Decoding Models Learn Paper • 2605.09992 • Published 6 days ago • 1
Attention Drift Collection Models trained as a part of the "Attention Drift: What Speculative Decoding Models Learn" paper, shared for reproducing experiments. • 14 items • Updated about 8 hours ago