How the community trained Gemma to "Think" with Tunix and TPUs- Google Developers Blog

This post details the outcomes of the Google Tunix Hackathon, where developers trained Gemma models to exhibit reasoning capabilities using Tunix and Kaggle TPUs. It highlights winning techniques such as Rubric-Based Reinforcement Learning (G-RaR), Pinocchio-1B's SFT → SimPO → GRPO pipeline, and IDEA-E Distillation with curriculum-guided GRPO. It also showcases domain-specific reasoning training for medical, chemistry, legal, and robotics applications, and provides resources for developers to train their own reasoning models.