Uncertainty Quantification and Decomposition for LLM-based Recommendation

Wonbin Kweon, Sanghwan Jang, SeongKu Kang, Hwanjo Yu · 2025

Read Uncertainty Quantification and Decomposition for LLM-based Recommendation by Wonbin Kweon, Sanghwan Jang, SeongKu Kang, Hwanjo Yu (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-2501-17630.txt.

Description

Wonbin Kweon, Sanghwan Jang, SeongKu Kang, and Hwanjo Yu measured uncertainty in two thousand twenty-five in recommendations made by large language models. Such models are now widely used to suggest items, yet they often sound sure even when their picks are shaky. Trustworthy use needs a way to grade that reliability.

Their frame estimates predictive uncertainty as a reliability score for each recommendation. It then splits that uncertainty into two parts: doubt about the recommendation itself, and doubt bred by the prompt. Tests show the score flags unreliable picks well. A prompting tweak that uses the split lowered uncertainty and improved recommendations. Code and weights are shared.

Catalog record

CREATORWonbin Kweon, Sanghwan Jang, SeongKu Kang, Hwanjo Yu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-17630

Browse the full library · About the Tower of Babel