Factuality on Demand: Controlling the Factuality–Informativeness Trade-off in Text Generation

Ziwei Gong, Yanda Chen, Julia Hirschberg, Chen Zhao, He He, Zhou Yu, Kathleen Mckeown · 2026

Read Factuality on Demand: Controlling the Factuality–Informativeness Trade-off in Text Generation by Ziwei Gong, Yanda Chen, Julia Hirschberg, Chen Zhao, He He, Zhou Yu, Kathleen Mckeown (2026) 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-2602-00848.txt.

Description

Ziwei Gong and colleagues introduce Factuality-Controlled Generation, a method letting users set truth rules with queries. Models face a trade-off: shorter but highly factual replies, or richer but potentially less accurate ones. Each use needs its own balance. They propose judging FCG on two fronts, respect for truth rules and response detail. They train models on the FCG task with synthetic data, and show this training greatly lifts both rule respect and detail.

Catalog record

CREATORZiwei Gong, Yanda Chen, Julia Hirschberg, Chen Zhao, He He, Zhou Yu, Kathleen Mckeown
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32419
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00848

Browse the full library · About the Tower of Babel