SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG

Retrieval-augmented generation (RAG) systems enhance large language models (LLMs) with external knowledge but are vulnerable to corpus poisoning and contamination attacks, which can compromise output integrity. Existing defenses often apply aggressive filtering, leading to unnecessary loss of valuable information and reduced reliability in generation. To address this problem, we propose a two-stage semantic filtering and conflict-free framework for trustworthy RAG. In the first stage, we perform a joint filter with semantic and cluster-based filtering which is guided by the Entity-intent-relation extractor (EIRE). EIRE extracts entities, latent objectives, and entity relations from both the user query and filtered documents, scores their semantic relevance, and selectively adds valuable documents into the clean retrieval database. In the second stage, we proposed an EIRE-guided conflict-aware filtering module, which analyzes semantic consistency between the query, candidate answers, and retrieved knowledge before final answer generation, filtering out internal and external contradictions that could mislead the model. Through this two-stage process, SeCon-RAG effectively preserves useful knowledge while mitigating conflict contamination, achieving significant improvements in both generation robustness and output trustworthiness. Extensive experiments across various LLMs and datasets demonstrate that the proposed SeCon-RAG markedly outperforms state-of-the-art defense methods.

Key Contributions

• EIRE (Entity-intent-relation extractor) that extracts entities, latent objectives, and entity relations from queries and documents to guide semantic relevance scoring and filtering
• Two-stage RAG defense: (1) joint semantic and cluster-based filtering to build a clean retrieval database, and (2) conflict-aware filtering that detects internal and external contradictions before final answer generation
• SeCon-RAG framework that outperforms state-of-the-art defenses across multiple LLMs and datasets on both robustness and generation quality under corpus poisoning

🛡️ Threat Analysis

Details

Similar Papers