ParamMute: Suppressing Knowledge-Critical FFNs for Faithful Retrieval-Augmented Generation

TL;DR

ParamMute is a novel method that suppresses specific internal FFNs in LLMs to improve the faithfulness of retrieval-augmented generation, reducing reliance on internal knowledge and aligning outputs with external evidence.

Contribution

This work identifies unfaithfulness-associated FFNs and introduces ParamMute, a framework that suppresses their activation to enhance LLM output faithfulness in RAG settings.

Findings

ParamMute significantly improves faithfulness on CoFaithfulQA.

It reduces reliance on internal parametric knowledge.

The approach outperforms existing methods on benchmark tests.

Abstract

Large language models (LLMs) integrated with retrieval-augmented generation (RAG) have improved factuality by grounding outputs in external evidence. However, they remain susceptible to unfaithful generation, where outputs contradict retrieved context despite its relevance and accuracy. Existing approaches aiming to improve faithfulness primarily focus on enhancing the utilization of external context, but often overlook the persistent influence of internal parametric knowledge during generation. In this work, we investigate the internal mechanisms behind unfaithful generation and identify a subset of mid-to-deep feed-forward networks (FFNs) that are disproportionately activated in such cases. Building on this insight, we propose Parametric Knowledge Muting through FFN Suppression (ParamMute), a framework that improves contextual faithfulness by suppressing the activation of unfaithfulness-associated FFNs and calibrating the model toward retrieved knowledge. To evaluate our approach, we introduce CoFaithfulQA, a benchmark specifically designed to evaluate faithfulness in scenarios where internal knowledge conflicts with accurate external evidence. Experimental results show that ParamMute significantly enhances faithfulness across both CoFaithfulQA and the established ConFiQA benchmark, achieving substantial reductions in reliance on parametric memory. These findings underscore the importance of mitigating internal knowledge dominance and provide a new direction for improving LLM trustworthiness in RAG. All codes are available at https://github.com/OpenBMB/ParamMute.