Genius Patrick

Train Your Own Encoder-Free VLM in $100

Building and scaling RL environments for LLM training

Explore on-policy distillation visualization for any model

TRL distillation for 100B+ teachers, 40x faster

Visualize synthetic‑data experiments as an interactive bookshelf

Who needs 1T parameters? Olympiad proofs with a 4B model

Explore LLM benchmark scores over time

The secrets to building world-class LLMs

A new open-source dataset for training VLMs

Read a comprehensive robot learning tutorial