Bridging External and Parametric Knowledge: Mitigating Hallucination of LLMs with Shared-Private Semantic Synergy in Dual-Stream Knowledge

TL;DR

This paper introduces DSSP-RAG, a novel framework that enhances retrieval-augmented generation in LLMs by distinguishing shared and private semantics, detecting hallucinations, and reducing external knowledge noise, thereby improving factual accuracy.

Contribution

It proposes a dual-stream semantic synergy framework with mixed-attention and an unsupervised hallucination detection method to better integrate external knowledge in LLMs.

Findings

DSSP-RAG outperforms strong baselines in experiments.

The mixed-attention mechanism effectively distinguishes shared and private semantics.

Unsupervised hallucination detection reduces unnecessary external knowledge incorporation.

Abstract

Retrieval-augmented generation (RAG) aims to mitigate the hallucination of Large Language Models (LLMs) by retrieving and incorporating relevant external knowledge into the generation process. However, the external knowledge may contain noise and conflict with the parametric knowledge of LLMs, leading to degraded performance. Current LLMs lack inherent mechanisms for resolving such conflicts. To fill this gap, we propose a Dual-Stream Knowledge-Augmented Framework for Shared-Private Semantic Synergy (DSSP-RAG). Central to it is the refinement of the traditional self-attention into a mixed-attention that distinguishes shared and private semantics for a controlled knowledge integration. An unsupervised hallucination detection method that captures the LLMs' intrinsic cognitive uncertainty ensures that external knowledge is introduced only when necessary. To reduce noise in external knowledge, an Energy Quotient (EQ), defined by attention difference matrices between task-aligned and task-misaligned layers, is proposed. Extensive experiments show that DSSP-RAG achieves a superior performance over strong baselines.