arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.20281cs.GTcs.AI
收藏

AI代理合谋的脆弱性

On the Fragility of AI Agent Collusion

National University of Singapore(新加坡国立大学);Boston University(波士顿大学)

Jussi Keppo, Yuze Li, Gerry Tsoukalas, Nuo Yuan

AI总结:

本文研究了AI代理合谋的脆弱性,发现异质性会减少合谋均衡。实验显示,耐心异质性和数据访问异质性削弱合谋,而模型大小差异反而稳定合谋,讨论了反垄断政策。

摘要:

近期研究表明,对称LLM代理的定价会导致算法合谋。本文显示,在真实部署中典型的异质性下,合谋是脆弱的。在简化的一次性定价模型中,耐心或数据访问的异质性会减少合谋均衡。使用开源LLM代理(总计超过2000个计算小时)的实验支持这些预测:耐心异质性将价格抬升从竞争水平以上的22%降至10%;不对称数据访问降至7%。增加竞争LLM的数量会打破合谋;交叉算法异质性(即LLM对Q学习代理)也是如此。但模型大小差异(例如32B vs. 14B权重)不会;它们产生领导者-跟随者动态,稳定合谋。我们讨论了反垄断影响,如限制数据共享的执法行动和促进算法多样性的政策。

英文摘要:

Recent work shows that pricing with symmetric LLM agents leads to algorithmic collusion. We show that collusion is fragile under the heterogeneity typical of real deployments. In a stylized repeated-pricing model, heterogeneity in patience or data access reduces the set of collusive equilibria. Experiments with open-source LLM agents (totaling over 2,000 compute hours) align with these predictions: patience heterogeneity reduces price lift from 22% to 10% above competitive levels; asymmetric data access, to 7%. Increasing the number of competing LLMs breaks up collusion; so does cross-algorithm heterogeneity, that is, setting LLMs against Q-learning agents. But model-size differences (e.g., 32B vs. 14B weights) do not; they generate leader-follower dynamics that stabilize collusion. We discuss antitrust implications, such as enforcement actions restricting data-sharing and policies promoting algorithmic diversity.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑