Hacker News new | past | comments | ask | show | jobs | submit

Attention at Constant Cost per Token via Symmetry-Aware Taylor Approximation

https://arxiv.org/abs/2602.00294
loading story #46887069
loading story #46888355
loading story #46888513
loading story #46886923
loading story #46890647
loading story #46887427
loading story #46887588
loading story #46886959
loading story #46886899
loading story #46887153
loading story #46890344
loading story #46887391
loading story #46886911
loading story #46888471
loading story #46887026