NLU-11: Week 9
This week, we we will have our two last examinable lectures on QA and RAG, and we have a non-examinable guest lecture on Fact Checking by Emily Allaway.
As a reminder, Coursework 2 is due on Friday 21 March at 12:00 noon. When submitting please make sure you *tag your partner* on Gradescope when you do that, otherwise ITO loses track of who did the assignment with whom
This week's lab is on prompting with GPT-3.5. It even includes a small competition on who can develop the best performing prompt on a small data set! Please note you will need an OpenAI key to run this lab; please obtain this before you attend your lab session, instructions can be found in the lab itself.
The solutions for last week's tutorial are now available:
And here is a preview of next week's tutorial. Please work through this before attending your tutorial group next week:
Tutorial 4: Practise Exam Questions
| Slides | Lecture | Course Content |
| Question Answering | 25 | Required Reading: Speech and Language Processing Ed. 3, Ch. 14 on RAG and QA Natural Questions: A Benchmark for Question Answering Research, https://aclanthology.org/Q19-1026/ Optional Reading: Bidirectional Attention Flow for Machine Comprehension, https://arxiv.org/abs/1611.01603 (BiDAF) SQuAD: 100,000+ Questions for Machine Comprehension of Text, https://arxiv.org/abs/1606.05250 (SQuAD) (Optional) Know What You Don’t Know: Unanswerable Questions for SQuAD, https://arxiv.org/abs/1806.03822 (SQuAD v2) Latent Retrieval for Weakly Supervised Open Domain Question Answering https://arxiv.org/abs/1906.00300 (Optional; worth a reading!) The Bitter Lesson: https://www.cs.utexas.edu/~eunsol/courses/data/bitter_lesson.pdf |
| Retrieval-Augmented Generation | 26 | Required Reading: ATLAS: Few-shot Learning with Retrieval Augmented Language Models, https://arxiv.org/abs/2208.03299 Optional Reading: Reliable, Adaptable, and Attributable Language Models with Retrieval, https://arxiv.org/abs/2403.03187 REALM: Retrieval-Augmented Language Model Pre-Training, https://arxiv.org/abs/2002.08909 Reading Wikipedia to Answer Open-Domain Questions, https://arxiv.org/abs/1704.00051 Question and Answer Test-Train Overlap in Open-Domain Question Answering Datasets, https://arxiv.org/abs/2008.02637 Challenges in Generalisation in Open Domain Question Answering, https://arxiv.org/abs/2109.01156 |
| Fact Checking | 27 | Guest Lecture by Emily Allaway Not examinable |