Stable-SPAM: How to Train in 4-Bit More Stably than 16-Bit Adam Paper • 2502.17055 • Published 4 days ago • 14
SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution Paper • 2502.18449 • Published 3 days ago • 54
SpargeAttn: Accurate Sparse Attention Accelerating Any Model Inference Paper • 2502.18137 • Published 3 days ago • 46
Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention Paper • 2502.11089 • Published 12 days ago • 134