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

2025/06/06 by Yi Sui, Chaozhuo Li, Sui, Yi +7 · 1 citation
Computer Science · #Advanced Graph Neural Networks #Bridging (networking) #Cognition #Computation and Language (cs.CL) #FOS: Computer and information sciences #Multimodal Machine Learning Applications #Noise (video) #Parametric statistics #Semantics (computer science) #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2506.06240

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2025/06/06 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

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.

Citations

• Loki's Dance of Illusions: A Comprehensive Survey of Hallucination in Large Language Models
• CAFE: Retrieval Head-based Coarse-to-Fine Information Seeking to Enhance Multi-Document QA Capability
• Synergizing RAG and Reasoning: A Systematic Review
• ReaRAG: Knowledge-guided Reasoning Enhances Factuality of Large Reasoning Models with Iterative Retrieval Augmented Generation
• Parameters vs. Context: Fine-Grained Control of Knowledge Reliance in Language Models
• Research on the Online Update Method for Retrieval-Augmented Generation (RAG) Model with Incremental Learning
• Knowledge Editing through Chain-of-Thought
• Qwen2.5 Technical Report
• Long-Context LLMs Meet RAG: Overcoming Challenges for Long Inputs in RAG
• Trustworthiness in Retrieval-Augmented Generation Systems: A Survey
• LLM Internal States Reveal Hallucination Risk Faced With a Query
• RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs
• Entropy-Based Decoding for Retrieval-Augmented Large Language Models
• To Believe or Not to Believe Your LLM
• Generalizing Knowledge Graph Embedding with Universal Orthogonal Parameterization
• ClashEval: Quantifying the tug-of-war between an LLM's internal prior and external evidence
• Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity
• Knowledge Conflicts for LLMs: A Survey
• TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space
• How Large Language Models Encode Context Knowledge? A Layer-Wise Probing Study
• Can Language Models Act as Knowledge Bases at Scale?
• BlendFilter: Advancing Retrieval-Augmented Large Language Models via Query Generation Blending and Knowledge Filtering
• Gated Linear Attention Transformers with Hardware-Efficient Training
• A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
• Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection
• Survey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-Specificity
• Llama 2: Open Foundation and Fine-Tuned Chat Models
• Lost in the Middle: How Language Models Use Long Contexts
• Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs
• Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy
• Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback
• Improving Language Models via Plug-and-Play Retrieval Feedback
• Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts
• In-Context Retrieval-Augmented Language Models
• Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Models
• Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions
• Reasoning with Language Model Prompting: A Survey
• Discovering Latent Knowledge in Language Models Without Supervision
• Mass-Editing Memory in a Transformer
• Language Models (Mostly) Know What They Know
• Teaching Models to Express Their Uncertainty in Words
• ASQA: Factoid Questions Meet Long-Form Answers
• Self-Consistency Improves Chain of Thought Reasoning in Language Models
• Synthetic Disinformation Attacks on Automated Fact Verification Systems
• HousE: Knowledge Graph Embedding with Householder Parameterization
• Attacking Open-domain Question Answering by Injecting Misinformation
• Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps
• Explainable Automated Fact-Checking for Public Health Claims
• HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
• TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for\n Reading Comprehension
• One SPACE to Rule Them All: Jointly Mitigating Factuality and Faithfulness Hallucinations in LLMs

Cited by

• Hallucination Detection via Internal States and Structured Reasoning Consistency in Large Language Models
• One SPACE to Rule Them All: Jointly Mitigating Factuality and Faithfulness Hallucinations in LLMs

Related

• Augmenting LLMs with Knowledge: A survey on hallucination prevention
• Rowen: Adaptive Retrieval-Augmented Generation for Hallucination Mitigation in LLMs
• ReDeEP: Detecting Hallucination in Retrieval-Augmented Generation via Mechanistic Interpretability
• MultiRAG: A Knowledge-guided Framework for Mitigating Hallucination in Multi-source Retrieval Augmented Generation
• LUMINA: Detecting Hallucinations in RAG System with Context-Knowledge Signals