Reinforcement Learning and Its Role in Modern LLMs

​In this talk, we will explore how reinforcement learning (RL) shapes modern language models, from foundational algorithms to cutting-edge techniques like RLHF and DeepSeek’s approach.

​Agenda:

• ​Introduction – Overview of RL and its role in LLMs

• ​Evolution of RL – From classical methods to AlphaGo & MuZero

• ​RL in LLMs – How reinforcement learning enhances language models

• ​RLHF & Alternatives – Comparing RLHF, DPO, and other strategies

• ​DeepSeek’s Approach – Using RL to improve LLM reasoning

• ​Future Perspectives – The next steps in optimising AI with RL

​Mariam Dundua has many years of experience and exceptional knowledge in data science, machine learning, and modern AI technologies. She will join us online.

​Link to the meeting: meet.google.com/cdf-ipoa-tsz