Researcher, ByteDance Inc.
1 paper at NeurIPS 2025
We propose a more expressive Transformer, which exceeding its original $TC^0$ expressiveness.