Programme 2025

Overview

The Oxford LLMs Programme is a five-day workshop that combines lectures, coding sessions, collaborative projects, and guest talks. Participants learn how large language models work, how to adapt them to new domains, how to evaluate them, and how to deploy them safely and fairly.

Daily Schedule

Where: Every day from 09:00–17:00 in the Large Lecture Room (Butler Room as backup).
Food: Fellows’ Dining Hall & Senior Common Room.

Monday 22 September: Fundamentals

• 09:00–09:15 — Morning coffee and biscuits
• 09:15–10:45 — Lecture: Mikhail Burtsev: Fundamentals of LLMs
• 10:45–11:00 — Coffee and biscuits
• 11:00–12:30 — Coding seminar: Ilya Boytsov: Fundamentals of LLMs
• 12:30–14:00 — Lunch
• 14:00–15:45 — Research: Introduction to project & outline of activities
• 15:45–16:00 — Coffee break
• 16:00–17:00 — Guest talk: Ray Duch: Synthetic Agents for Benchmarking Experiments
• 19:00–21:00 — Social

Tuesday 23 September: LLM Evaluation

• 09:00–09:15 — Morning coffee and biscuits
• 09:15–10:45 — Coding seminar: Emeli Dral: LLM Evaluation
• 10:45–11:00 — Coffee and biscuits
• 11:00–12:30 — Lecture: Tatiana Shavrina: LLM Evaluation
• 12:30–14:00 — Lunch
• 14:00–15:45 — Research
• 15:45–16:00 — Coffee break
• 16:00–17:00 — Guest talk (online): Joan Timoneda: Using LLMs to Identify Overt and Covert Racism in Spanish

Wednesday 24 September: AI Agents & LLM Fine-Tuning

• 09:00–09:15 — Morning coffee and biscuits
• 09:15–10:45 — Lecture: Grigory Sapunov: AI Agents
• 10:45–11:00 — Coffee and biscuits
• 11:00–12:30 — Coding seminar: Sergei Skvortsov: LLM Fine-Tuning
• 12:30–14:00 — Lunch
• 14:00–15:00 — Guest talk: Brandon Stewart (with Metrics & Models seminar series)
• 15:00–15:30 — Coffee break
• 15:30–17:00 — Research

Thursday 25 September: Research & Oxford Tour

• 09:00–09:15 — Morning coffee and biscuits
• 09:15–10:30 — Research
• 10:30–13:00 — Oxford Tour
• 13:15–14:15 — Lunch
• 14:15-15:45 — Research
• 15:45–16:00 — Coffee break
• 16:00–17:00 — Guest talk (online): Kosuke Imai: GenAI-Powered Inference

Friday 26 September: Research, Safety, Outro

• 09:00–09:15 — Morning coffee and biscuits
• 09:15–10:45 — Lecture: Denis Volk: AI Safety
• 10:45–11:00 — Coffee and biscuits
• 11:00–12:30 — Research
• 12:30–14:00 — Lunch
• 14:00–15:45 — Research Presentations by participants
• 15:45–16:00 — Coffee break
• 16:00–17:00 — Outro talk: Charles Rahal: On the Unkowable Limits to Prediction