Teaching syllabus
2026-2027 · Course index
Three online sessions in the autumn, then four half days on site in January, each pairing a lecture with a lab. Topics run from the Transformer architecture through training, alignment and reasoning, up to agents.
- 11 sessions
- 7 lectures
- 4 labs
- 24 hours
- 01Mon 26 Oct12:15 – 13:45TransformerWe start with the main NLP tasks and the architectures used before Transformers. We then look at why recurrent networks struggle with long dependencies, how attention works, and how it leads to the Transformer architecture.CourseOnlineQuiz readyHandout to come
- 02Wed 28 Oct12:15 – 13:45Transformer-based models and tricksAttention on its own carries no information about the order of tokens. We look at how models represent position and how the approaches evolved up to RoPE.CourseOnlineQuiz readyHandout to come
- 03Thu 03 Dec12:15 – 13:45Large Language ModelsThis session covers large language models: their architecture, text generation, prompting, and the main techniques used to speed up inference.CourseOnlineQuiz readyHandout to come
- 04Thu 14 Jan08:45 – 10:15LLM trainingThis session starts with how language models are trained: pretraining, the cost in compute and memory, distributed training and the main optimisations.CourseOn siteQuiz readyHandout to come
- 05Thu 14 Jan10:30 – 12:00To be updatedTo be updatedLabOn siteHandout to come
- 06Thu 14 Jan14:00 – 15:30LLM tuningA pretrained model can produce text, but that does not mean it follows an instruction correctly or always gives a suitable answer.CourseOn siteQuiz readyHandout to come
- 07Thu 14 Jan15:45 – 17:15To be updatedTo be updatedLabOn siteHandout to come
- 08Fri 15 Jan14:00 – 15:30LLM reasoningThis session covers models trained to carry out longer reasoning on certain tasks.CourseOn siteQuiz readyHandout to come
- 09Fri 15 Jan15:45 – 17:15To be updatedTo be updatedLabOn siteHandout to come
- 10Sat 16 Jan08:45 – 10:15Agentic LLMsThe last lecture brings together the main building blocks used in the project: information retrieval, search over a document base, tool calling and agents.CourseOn siteQuiz readyHandout to come
- 11Sat 16 Jan10:30 – 12:00To be updatedTo be updatedLabOn siteHandout to come