Reading up on GRPO
2 deep · digging since nov 27, 25
- Tiny World Model - moomou
The author built a tiny VAE‑RNN world model for a Dino clone, trained its controller with GRPO, and showed each part needs its own honest metric.
- AI infrastructure in the "Era of experience"
As open-source LLMs commoditize due to fierce Chinese competition, companies can build defensible moats through custom RL models trained on proprietary environments using LoRA and GRPO.