Blog

State of RL for reasoning LLMs

March 15, 2026 · 26 min read

Evolution of reinforcement learning for reasoning LLMs

LoRA initialization

December 15, 2025 · 8 min read

Showing that $0\neq 0$

nanochat's gpt.py

November 04, 2025 · 24 min read

Step-by-step exploration of nanochat's model

DPO is SFT

September 07, 2025 · 16 min read

Close examination of gradients of DPO and SFT