Alessio Frateily's Highlights on 'Large language model training: how three training phases shape LLMs' | Glasp