Information Retrieval for AI
KAIST AI, Autumn 2026. Mondays and Wednesdays, 09:00 KST, on Zoom. No exam.
Next
Session 12, Mon 19 Oct. Attention as retrieval. Attention as a lookup, retrieval-head evidence and the limits of the analogy.
Mon 19 Oct, 09:00 in Asia/Seoul - 5 days away
Reading this as a visitor
The syllabus and every delivered lecture deck are open to anyone and need no account. Everything else needs the email you are enrolled with.
The decks
1 - Mon 31 Aug Why retrieval now 2 - Wed 02 Sep The inverted index 3 - Mon 07 Sep Index construction and ranking foundations 4 - Wed 09 Sep BM25: finish probabilistic ranking 5 - Mon 14 Sep Evaluation and the changing goals of retrieval benchmarks 6 - Wed 16 Sep Coarse retrieval and reranking 7 - Mon 28 Sep LLM embeddings and reranking
8 - Wed 30 Sep BEIR
9 - Wed 07 Oct Hard queries: reasoning and ambiguity 10 - Mon 12 Oct Approximate nearest neighbour search and search systems 11 - Wed 14 Oct Project workshop 1 - guided modification
12 - Mon 19 Oct Attention as retrieval
13 - Wed 21 Oct LIMIT
14 - Mon 26 Oct Memory as an architectural axis
15 - Wed 28 Oct Retrieval Heads
16 - Mon 02 Nov Memory learned at test time
17 - Wed 04 Nov Engram
18 - Mon 09 Nov Language models with an index attached
19 - Wed 11 Nov Project workshop 2 - exploration
20 - Mon 16 Nov Retrieval-augmented generation
21 - Wed 18 Nov Titans
22 - Mon 23 Nov Agentic retrieval and deep research
23 - Wed 25 Nov Agent memory: write, manage, read
24 - Mon 30 Nov Learning what to remember
25 - Wed 02 Dec Agentic Memory
26 - Mon 07 Dec Evaluating agent memory
27 - Wed 09 Dec Evaluating Memory in LLM Agents
28 - Mon 14 Dec Privacy, personalisation and unlearning
29 - Wed 16 Dec Project workshop 3 - highlights
lecture discussion workshop 9 of 29 delivered - a delivered deck needs no account