Teaming LLMs to Detect and Mitigate Hallucinations

TL;DR

This paper introduces a consortium consistency approach that combines responses from multiple diverse LLMs to improve hallucination detection and mitigation, achieving better performance and lower inference costs than single-model methods.

Contribution

It extends single-model consistency techniques to multi-model ensembles, demonstrating significant improvements in hallucination mitigation across various LLMs and conditions.

Findings

Enhanced hallucination detection accuracy with multi-LLM ensembles

Reduced inference costs compared to single-model approaches

Effective across diverse model architectures and training data

Abstract

Recent work has demonstrated state-of-the-art results in large language model (LLM) hallucination detection and mitigation through consistency-based approaches which involve aggregating multiple responses sampled from a single LLM for a given prompt. These approaches help offset limitations stemming from the imperfect data on which LLMs are trained, which includes biases and under-representation of information required at deployment time among other limitations which can lead to hallucinations. We show that extending these single-model consistency methods to combine responses from multiple LLMs with different training data, training schemes and model architectures can result in substantial further improvements in hallucination detection and mitigation capabilities beyond their single-model consistency counterparts. We evaluate this "consortium consistency" approach across many model teams from a pool of 15 LLMs and explore under what conditions it is beneficial to team together different LLMs in this manner. Further, we show that these performance improvements often come with reduced inference costs, offsetting a significant drawback with single-model consistency methods.