Document Attribution: Examining Citation Relationships using Large Language Models

Vipula Rawte, Ryan Rossi, Franck Dernoncourt, Nedim Lipka · 2025

Read Document Attribution: Examining Citation Relationships using Large Language Models by Vipula Rawte, Ryan Rossi, Franck Dernoncourt, Nedim Lipka (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-06324.txt.

Description

Vipula Rawte and three co-authors study how language models cite documents in this 2025 paper. Models increasingly answer from supplied documents, and their citations must actually support the text. The team proposes two checks. The first treats attribution as a plain entailment question and asks a model whether the cited passage implies the claim. On a standard benchmark, that zero shot method edged past the best baselines on both in domain and out of domain sets.

The second check reads the model's own attention. Using a small model, attention based scores beat the baseline at nearly every layer. Two layers ranges lagged, a detail the authors report openly. The work matters because fluent answers with fake support are worse than no citation at all. Cheap entailment checks could run as a live filter on model output.

Catalog record

CREATORVipula Rawte, Ryan Rossi, Franck Dernoncourt, Nedim Lipka
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE17 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-06324

Browse the full library · About the Tower of Babel