Improving Contextual Faithfulness of Large Language Models via Retrieval Heads-Induced Optimization

TL;DR

This paper introduces RHIO, a novel training framework that enhances the contextual faithfulness of retrieval-augmented large language models by discriminating between faithful and unfaithful outputs, and also presents GroundBench, a new benchmark for evaluation.

Contribution

The paper proposes RHIO, a method that explicitly teaches LLMs to distinguish faithful from unfaithful generations using retrieval head masking and contrastive decoding, and introduces GroundBench for evaluation.

Findings

RHIO significantly improves faithfulness in LLMs.

RHIO outperforms GPT-4o on GroundBench.

GroundBench provides a comprehensive evaluation benchmark.

Abstract

Ensuring contextual faithfulness in retrieval-augmented large language models (LLMs) is crucial for building trustworthy information-seeking systems, particularly in long-form question-answering (LFQA) scenarios. In this work, we identify a salient correlation between LFQA faithfulness and retrieval heads, a set of attention heads responsible for retrieving contextual information. Leveraging this insight, we propose RHIO, a framework designed to teach LLMs to explicitly discriminate between faithful and unfaithful generations. RHIO first augments unfaithful samples that simulate realistic model-intrinsic errors by selectively masking retrieval heads. Then, these samples are incorporated into joint training, enabling the model to distinguish unfaithful outputs from faithful ones conditioned on control tokens. Furthermore, these control tokens are leveraged to self-induce contrastive outputs, amplifying their difference through contrastive decoding. Additionally, to facilitate the evaluation of contextual faithfulness, we also introduce GroundBench, a comprehensive benchmark compiled from five existing LFQA datasets. Extensive experimental results on GroundBench demonstrate that RHIO significantly improves faithfulness, even outperforming GPT-4o.