ValueFlow : Measuring the Propagation of Value Perturbations in Multi-Agent LLM Systems

Jinnuo Liu, Chuke Liu, Hua Shen · 2026

Read ValueFlow : Measuring the Propagation of Value Perturbations in Multi-Agent LLM Systems by Jinnuo Liu, Chuke Liu, Hua Shen (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-2602-08567.txt.

Description

Value alignment is usually tested on single models, but deployed systems increasingly chain agents that react to each other's outputs. Jinnuo Liu and two co-authors ask how a perturbed value in one agent spreads through such a system. Their arXiv paper introduces ValueFlow, a framework built on a 56-value dataset drawn from the Schwartz survey.

Drift is split into agent-level response and system-level structure, tracked by two measures: eta-susceptibility for single agents and system susceptibility for final outputs. Tests across values, backbones, personas, and topologies show susceptibility varies sharply and is shaped by interaction structure. Readers learn that alignment in multi-agent systems is a property of the whole network, not just its parts.

Catalog record

CREATORJinnuo Liu, Chuke Liu, Hua Shen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-08567

Browse the full library · About the Tower of Babel