SMILES-2025: From Markov Chains to Transformers: The Evolution of LLM | Irina Piontkovskaya

The opening talk in the “LLM from A to Z” series at SMILES‑2025. Irina Piontkovskaya, Head of NLP Research at RRI and author of 60 NLP papers, retraces three decades of language‑model progress: * n‑gram statistics and Markov chains, * the rise of recurrent nets and attention, * today’s giant transformers powering GPT, Gemini and RuLLM. Drawing on 13 years of personal research, she explains the paradigm shift from statistical to neural methods, the key breakthroughs that enabled modern LLMs, and what practitioners should know before fine‑tuning large models. The lecture lays the groundwork for all subsequent LLM discussions at SMILES‑2025