Accelerating RL Post-Training Rollouts via System-Integrated Speculative Decoding Paper • 2604.26779 • Published 8 days ago • 13
TriAttention: Efficient Long Reasoning with Trigonometric KV Compression Paper • 2604.04921 • Published Apr 6 • 112
Synthetic Sandbox for Training Machine Learning Engineering Agents Paper • 2604.04872 • Published Apr 6 • 14