Diversity Collapse in Multi-Agent LLM Systems: Structural Coupling and Collective Failure in Open-Ended Idea Generation

Nuo Chen 1 Yicheng Tong 1 Yuzhe Yang 2 Yufei He 1 Xueyi Zhang 2 Qingyun Zou 1 Qian Wang 1 Bingsheng He 1 1 National University of Singapore 2 The Chinese University of Hong Kong, Shenzhen · 2026

Read Diversity Collapse in Multi-Agent LLM Systems: Structural Coupling and Collective Failure in Open-Ended Idea Generation by Nuo Chen 1 Yicheng Tong 1 Yuzhe Yang 2 Yufei He 1 Xueyi Zhang 2 Qingyun Zou 1 Qian Wang 1 Bingsheng He 1 1 National University of Singapore 2 The Chinese University of Hong Kong, Shenzhen (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-2604-18005.txt.

Description

Nuo Chen and seven colleagues study diversity in multi-agent systems for open-ended idea generation across model, agent, and system levels. At model level, stronger aligned models give less marginal diversity despite higher sample quality, a compute efficiency paradox. Authority dynamics suppress meaning diversity relative to junior groups, while larger groups yield less return and dense talk speeds early convergence. The authors attribute this diversity collapse mainly to interaction shape, not weak models, urging designs that preserve independence and disagreement.

Catalog record

CREATORNuo Chen 1 Yicheng Tong 1 Yuzhe Yang 2 Yufei He 1 Xueyi Zhang 2 Qingyun Zou 1 Qian Wang 1 Bingsheng He 1 1 National University of Singapore 2 The Chinese University of Hong Kong, Shenzhen
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE180605
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.18005

Browse the full library · About the Tower of Babel