Seijin Kobayashi

MesaNet: Sequence Modeling by Locally Optimal Test-Time Training

Attention as a Hypernetwork

Multi-agent cooperation through learning-aware policy gradients

Learning Randomized Algorithms with Transformers