arxivcs.AIcs.GT2026-07-05
Decentralized Aggregation of LLM Predictions via Wagering Mechanisms
Yuhong Luo, David M. Pennock, Xintong Wang
It is increasingly common to aggregate predictions from multiple LLMs, each with domain expertise or access to private tools and data, to improve collective prediction performance. In decentralized settings, aggregation weights need to be determined without access to models' priv…