TokDrift: When LLM Speaks in Subwords but Code Speaks in Grammar Paper • 2510.14972 • Published 20 days ago • 29
LightMem: Lightweight and Efficient Memory-Augmented Generation Paper • 2510.18866 • Published 15 days ago • 107
Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning Paper • 2510.19338 • Published 14 days ago • 110
Running on CPU Upgrade 1.43k 1.43k The Smol Training Playbook: The Secrets to Building World-Class LLMs 📝
