You Only Align Once: Propagating Cooperative Behaviors in Multi-Agent Systems through Seed Agents

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hsing, Nicole, Zheng, Asuka Yuxi, Zhao, Yi, Tu, Haoqin, Huang, Jen-Tse
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914606596751360
author Hsing, Nicole
Zheng, Asuka Yuxi
Zhao, Yi
Tu, Haoqin
Huang, Jen-Tse
author_facet Hsing, Nicole
Zheng, Asuka Yuxi
Zhao, Yi
Tu, Haoqin
Huang, Jen-Tse
contents Ensuring agent behaviors in distributed open multi-agent systems remains challenging, especially as populations grow and unaligned agents may exist. We show that a single aligned agent can propagate cooperative behaviors to untrained agents purely through natural language interaction, a phenomenon we term Alignment Propagation. We study this in the Red-Black Game, a team-based iterated Prisoner's Dilemma in which teammates deliberate and vote to determine their team's collective action. By distilling the cooperative reasoning and persuasive dialogues of a teacher model into a Qwen-3-14B, we obtain a seed agent that, when placed among four untrained teammates, doubles the cooperation rate from 24.8% to 62.2%, outperforming the teacher model and a vanilla Gemini-3.1-Pro. Remarkably, a seed trained exclusively on the RedBlack Game transfers zero-shot to Sugarscape, a spatially grounded survival simulation with pairwise trading, achieving a 91.5% trade success rate versus a 21.6% baseline. Our results reframe multi-agent alignment from an exhaustive per-agent training problem to a scalable social capability that can be engineered through strategic seed placement.
format Preprint
id arxiv_https___arxiv_org_abs_2605_27586
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle You Only Align Once: Propagating Cooperative Behaviors in Multi-Agent Systems through Seed Agents
Hsing, Nicole
Zheng, Asuka Yuxi
Zhao, Yi
Tu, Haoqin
Huang, Jen-Tse
Multiagent Systems
Computation and Language
Ensuring agent behaviors in distributed open multi-agent systems remains challenging, especially as populations grow and unaligned agents may exist. We show that a single aligned agent can propagate cooperative behaviors to untrained agents purely through natural language interaction, a phenomenon we term Alignment Propagation. We study this in the Red-Black Game, a team-based iterated Prisoner's Dilemma in which teammates deliberate and vote to determine their team's collective action. By distilling the cooperative reasoning and persuasive dialogues of a teacher model into a Qwen-3-14B, we obtain a seed agent that, when placed among four untrained teammates, doubles the cooperation rate from 24.8% to 62.2%, outperforming the teacher model and a vanilla Gemini-3.1-Pro. Remarkably, a seed trained exclusively on the RedBlack Game transfers zero-shot to Sugarscape, a spatially grounded survival simulation with pairwise trading, achieving a 91.5% trade success rate versus a 21.6% baseline. Our results reframe multi-agent alignment from an exhaustive per-agent training problem to a scalable social capability that can be engineered through strategic seed placement.
title You Only Align Once: Propagating Cooperative Behaviors in Multi-Agent Systems through Seed Agents
topic Multiagent Systems
Computation and Language
url https://arxiv.org/abs/2605.27586