Ideas in Inference-time Scaling can Benefit Generative Pre-training Algorithms

DiffM
Re-assign community