DnDScore: Decontextualization and Decomposition for Factuality Verification in Long-Form Text Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wanner, Miriam, Van Durme, Benjamin, Dredze, Mark
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929635495772160
author Wanner, Miriam
Van Durme, Benjamin
Dredze, Mark
author_facet Wanner, Miriam
Van Durme, Benjamin
Dredze, Mark
contents The decompose-then-verify strategy for verification of Large Language Model (LLM) generations decomposes claims that are then independently verified. Decontextualization augments text (claims) to ensure it can be verified outside of the original context, enabling reliable verification. While decomposition and decontextualization have been explored independently, their interactions in a complete system have not been investigated. Their conflicting purposes can create tensions: decomposition isolates atomic facts while decontextualization inserts relevant information. Furthermore, a decontextualized subclaim presents a challenge to the verification step: what part of the augmented text should be verified as it now contains multiple atomic facts? We conduct an evaluation of different decomposition, decontextualization, and verification strategies and find that the choice of strategy matters in the resulting factuality scores. Additionally, we introduce DnDScore, a decontextualization aware verification method which validates subclaims in the context of contextual information.
format Preprint
id arxiv_https___arxiv_org_abs_2412_13175
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle DnDScore: Decontextualization and Decomposition for Factuality Verification in Long-Form Text Generation
Wanner, Miriam
Van Durme, Benjamin
Dredze, Mark
Computation and Language
The decompose-then-verify strategy for verification of Large Language Model (LLM) generations decomposes claims that are then independently verified. Decontextualization augments text (claims) to ensure it can be verified outside of the original context, enabling reliable verification. While decomposition and decontextualization have been explored independently, their interactions in a complete system have not been investigated. Their conflicting purposes can create tensions: decomposition isolates atomic facts while decontextualization inserts relevant information. Furthermore, a decontextualized subclaim presents a challenge to the verification step: what part of the augmented text should be verified as it now contains multiple atomic facts? We conduct an evaluation of different decomposition, decontextualization, and verification strategies and find that the choice of strategy matters in the resulting factuality scores. Additionally, we introduce DnDScore, a decontextualization aware verification method which validates subclaims in the context of contextual information.
title DnDScore: Decontextualization and Decomposition for Factuality Verification in Long-Form Text Generation
topic Computation and Language
url https://arxiv.org/abs/2412.13175