learn.ifs.name

Information Retrieval and Retrieval-Augmented Generation

Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 11

7 cards in “Introduction” · whole deck

Introduction

  1. What are factoid questions?
    Information needs that can be met with facts expressed in short texts, e.g. “Where is the Louvre Museum located?”.
  2. How can an instruction-tuned LLM answer factoid questions without retrieval, and where is that knowledge stored?
    By conditional generation from the question: it has seen the facts in pretraining. Such factual knowledge seems to be stored in the large feedforward layers of the transformer.
  3. What is a hallucination?
    A response that is not faithful to the facts of the world: the model makes up a reasonable-sounding answer. LLMs sometimes do this even when the correct facts are stored in their parameters, apparently when the feedforward layers fail to recall them.
  4. What does it mean for a system to be calibrated, and why does it matter for hallucination?
    Its confidence in an answer correlates highly with the probability that the answer is correct. A calibrated system could hedge when likely wrong; LLMs are poorly calibrated and often give wrong answers with complete certainty.
  5. What three problems arise from answering questions using only an LLM’s parameters?
    Hallucination (often wrong, hard to detect), no access to proprietary data (personal email, medical records, company documents), and static knowledge (no information after the model was trained).
  6. What is retrieval-augmented generation (RAG)?
    Using information retrieval to fetch documents likely to contain relevant information, then having an LLM generate an answer conditioned on those documents and the question.
  7. What are the benefits of grounding answers in retrieved documents?
    Answers are based on facts from a curated collection, can be shown together with their source passages (so users can check them), and retrieval works over any proprietary data.

Source