Large Language Models: Notes & Past Papers
Weeks 1 to 4 of the DA5004 paper series in one place: rigorous notes with runnable code and figures, beginner-friendly exam guides, and the past-year questions with scanned stems and worked answers. Use Ask AI on any page, section or selected passage to get an ELI15 explanation.
Start with a question you want to answer
Find every question on a topic
Scaled dot-product attention, masked self-attention, parameter counting, beam search, top-p sampling, BERT masking and more. Each topic links to every paper that tested it, with the worked answer; filter by year or hide repeated questions.
Browse past-paper topics →What to learn first, and why
Topics ranked by how often and how heavily they were tested across the routed papers.
Open the exam study route →New here? Start in plain language
Weekly notes, exam guides and past papers
From Sequence Models to Transformer Encoders
73 past-paper questionsTransformer Decoding, Position, Normalization, and Training
65 past-paper questionsLanguage Modelling, GPT-1, and Transfer
36 past-paper questionsDecoding Strategies and Bidirectional Language Representation
67 past-paper questionsReference
- Paper and question index: all 26 exam papers and how every question was routed.
- Archive: later-week questions: questions outside Weeks 1 to 4, with their routing reason and a link to the extracted paper.
- Knowledge map: how the concepts connect.
- How this was checked: verification of routing, question banks and links.
Worked answers are independently derived study aids, not official answer keys. Where a source was ambiguous or its extracted text disagreed with the scan, the page says so beside the question. The DA5004 course code is used only to name the paper series.