Attention Proven d² Not n² Problem
Anonymous Korean paper proves Attention's optimization landscape is d²-dimensional, not n², via d² Pullback Theorem. Proposes CSQ Attention with quadratic kernel for O(nd³) complexity. Calls for expert verification to rethink Transformers.






