One SPACE to Rule Them All: Jointly Mitigating Factuality and Faithfulness Hallucinations in LLMs

TL;DR

This paper introduces SPACE, a unified method that identifies and edits shared activation subspaces in LLMs to simultaneously reduce factuality and faithfulness hallucinations, improving model reliability.

Contribution

It reveals shared neural subspaces for hallucination types and proposes a novel editing framework to mitigate both issues concurrently.

Findings

SPACE outperforms existing methods on benchmark datasets.

Shared subspaces are crucial for hallucination mitigation.

Joint editing reduces both factuality and faithfulness hallucinations.

Abstract

LLMs have demonstrated unprecedented capabilities in natural language processing, yet their practical deployment remains hindered by persistent factuality and faithfulness hallucinations. While existing methods address these hallucination types independently, they inadvertently induce performance trade-offs, as interventions targeting one type often exacerbate the other. Through empirical and theoretical analysis of activation space dynamics in LLMs, we reveal that these hallucination categories share overlapping subspaces within neural representations, presenting an opportunity for concurrent mitigation. To harness this insight, we propose SPACE, a unified framework that jointly enhances factuality and faithfulness by editing shared activation subspaces. SPACE establishes a geometric foundation for shared subspace existence through dual-task feature modeling, then identifies and edits these subspaces via a hybrid probe strategy combining spectral clustering and attention head saliency scoring. Experimental results across multiple benchmark datasets demonstrate the superiority of our approach.