Aggregating probability estimates from 15 LLMs with a learned linear model beats individual models and classical voting rules on clean questions, and the apparent cloud-vs-local capability gap largely vanishes under contamination-free evaluation.
Title resolution pending
1 Pith paper cite this work, alongside 2 external citations. Polarity classification is still indexing.
1
Pith paper citing it
2
external citations · OpenAlex
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Wisdom of LLM Crowds: Aggregation and Contamination in Language Model Ensembles
Aggregating probability estimates from 15 LLMs with a learned linear model beats individual models and classical voting rules on clean questions, and the apparent cloud-vs-local capability gap largely vanishes under contamination-free evaluation.