The paper that kick-started LLM development is attention_is_all_you_need from Google, published in June 2017. It introduced Transformer architecture and “self-attention” mechanism.