Research map / Large language models / Retrieval and knowledge
Safe RAG by RAG: Untying the Bell That RAG Rang with the RAG Hand
Xun Liang, Mengwei Wang, Yuefeng Ma, Simin Niu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add RAG2RAG Detective-Judge expert to the main RAG | main RAG · RSB BPI attack | ASR: 0.94 -> 0.00 (-0.94) | Improved |
| add RAG2RAG expert module | main RAG · RSB WPI attack | ACC: +0.83 | Improved |
| add RAG2RAG expert module | main RAG · SafeRAG Inter-Context Conflict | ASR: -0.19 (0.07) | Improved |
| use RAG2RAG instead of TrustRAG on BPRAG attack | RSB BPRAG · RSB BPRAG | ASR: 0.05 -> 0.15 | Hurt |
| add the main RAG retriever back to the Detective/Judge pipeline (ablation) | RAG2RAG · RSB | ACC: 0.64 -> 0.81 | Improved |
| add the main RAG retriever back (ablation) | RAG2RAG · RSB | ASR: 0.27 -> 0.27 | No clear effect |
| keep the main RAG generator instead of Judge-as-generator (ablation) | RAG2RAG · RSB | ACC: 0.73 -> 0.81 | Improved |
| keep the main RAG pipeline instead of Judge-as-generator (ablation) | RAG2RAG · SafeRAG | F1: 0.27 -> 0.40 | Improved |
| plug RAG2RAG into different RAG variants (AAR, SuRe, Adaptive-RAG, IRCoT, FLARE, RQRAG) | advanced RAG frameworks · RSB | ASR | Improved |
| use Detective top-K retrieval instead of TrustRAG-style clustering | RAG defense · RSB time cost | latency | Improved |
Closest papers by meaning
- On the Vulnerability of Applying Retrieval-Augmented Generation within Knowledge-Intensive Application Domains (ICML 2025 · similarity 0.87)
- Follow My Instruction and Spill the Beans: Scalable Data Extraction from Retrieval-Augmented Generation Systems (ICLR 2025 · similarity 0.87)
- ReliabilityRAG: Effective and Provably Robust Defense for RAG-based Web-Search (NeurIPS 2025 · similarity 0.86)
- SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG (NeurIPS 2025 · similarity 0.85)
- RAGChecker: A Fine-grained Framework for Diagnosing Retrieval-Augmented Generation (NeurIPS 2024 · similarity 0.85)
- RAGGED: Towards Informed Design of Scalable and Stable RAG Systems (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
