Recursive Sparse Attention: Million-Token Extrapolation at Logarithmic Cost

4 points | by matt_d 8 hours ago

No comments yet.