Stepsize anything: A unified learning rate schedule for budgeted-iteration training Paper • 2505.24452 • Published May 30 • 5
Stepsize anything: A unified learning rate schedule for budgeted-iteration training Paper • 2505.24452 • Published May 30 • 5
Stepsize anything: A unified learning rate schedule for budgeted-iteration training Paper • 2505.24452 • Published May 30 • 5 • 2
Seed1.5-Thinking: Advancing Superb Reasoning Models with Reinforcement Learning Paper • 2504.13914 • Published Apr 10 • 4
Expert Race: A Flexible Routing Strategy for Scaling Diffusion Transformer with Mixture of Experts Paper • 2503.16057 • Published Mar 20 • 14
Frac-Connections: Fractional Extension of Hyper-Connections Paper • 2503.14125 • Published Mar 18 • 21
Frac-Connections: Fractional Extension of Hyper-Connections Paper • 2503.14125 • Published Mar 18 • 21
HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization Paper • 2503.04598 • Published Mar 6 • 20 • 8
HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization Paper • 2503.04598 • Published Mar 6 • 20
HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization Paper • 2503.04598 • Published Mar 6 • 20 • 8