When One LLM Drools, Multi-LLM Collaboration Rules

Shangbin Feng, Wenxuan Ding, Alisa Liu, Zifeng Wang, Weijia Shi, Yike Wang, Shannon Zejiang Shen, Xiaochuang Han, Hunter Lang, Chen-Yu Lee, Tomas Pfister, Yejin Choi, Yulia Tsvetkov · 2025

Read When One LLM Drools, Multi-LLM Collaboration Rules by Shangbin Feng, Wenxuan Ding, Alisa Liu, Zifeng Wang, Weijia Shi, Yike Wang, Shannon Zejiang Shen, Xiaochuang Han, Hunter Lang, Chen-Yu Lee, Tomas Pfister, Yejin Choi, Yulia Tsvetkov (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-2502-04506.txt.

Description

Shangbin Feng and twelve co-authors argue, in a position paper, that one large language model is often not enough. Real tasks are complex, contextual, and subjective. A single model, they claim, cannot represent the full spread of human data, skills, and viewpoints, and no amount of extra training fully patches those gaps.

The authors organize ways for models to collaborate into a ladder: through public APIs, through exchanged text, through shared scores, and through merged weights. Collaboration, they argue, buys reliability, wider access, and room for plural views, all weak points of the solo model. They also catalog where current team-ups fall short. Their closing vision is compositional intelligence: systems assembled from many models that check and complete each other.

Catalog record

CREATORShangbin Feng, Wenxuan Ding, Alisa Liu, Zifeng Wang, Weijia Shi, Yike Wang, Shannon Zejiang Shen, Xiaochuang Han, Hunter Lang, Chen-Yu Lee, Tomas Pfister, Yejin Choi, Yulia Tsvetkov
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-04506

Browse the full library · About the Tower of Babel