Towards Effective Extraction and Evaluation of Factual Claims

Dasha Metropolitansky, Jonathan Larson · 2025

Read Towards Effective Extraction and Evaluation of Factual Claims by Dasha Metropolitansky, Jonathan Larson (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-2502-10855.txt.

Description

Dasha Metropolitansky and Jonathan Larson improve how fact-checkers pull claims from machine text. Their 2025 paper tackles a weak link in checking long AI answers. Checkers first split text into simple claims that can be verified one by one. Sloppy claims quietly poison the verdicts that follow.

The team proposes a shared yardstick for judging claim extraction. Their automated checks measure coverage and how well claims stand alone. A new extractor, Claimify, works under this yardstick. It handles ambiguity and only pulls a claim when the reading is clear. Under the new tests, Claimify beat existing extractors. Shared measures let future methods compete fairly.

Catalog record

CREATORDasha Metropolitansky, Jonathan Larson
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE176 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-10855

Browse the full library · About the Tower of Babel