Lost in the Middle: How Language Models Use Long Contexts
Liu et al., 2024, Lost in the Middle (Transactions of ACL, vol. 12)
Subscriber deck — free during early access
1 cards in “2.1 Experimental Setup” · whole deck
2.1 Experimental Setup
- How is the multi-document QA task built?Each NaturalQuestions-Open question comes with \(k\) Wikipedia passages: exactly one contains the answer, and \(k - 1\) are distractors retrieved by Contriever as most relevant but answer-free. Context length is varied via \(k\) (10, 20, 30), position by reordering the passages.
- Liu et al., 2024, Lost in the Middle (Transactions of ACL, vol. 12), §2.1, p. 159
Source
- Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni and Percy Liang. 2024. Lost in the Middle: How Language Models Use Long Contexts. In Transactions of the Association for Computational Linguistics, vol. 12, pp. 157–173. License: CC BY 4.0. Flashcards adapted from this work. doi:10.1162/tacl_a_00638 https://aclanthology.org/2024.tacl-1.9/