Retrieval-augmented generation
TrustRAG
TrustRAG: Enhancing Robustness and Trustworthiness in Retrieval-Augmented Generation
Superseded — cited as a baseline and beaten by newer methods
4 papers critique it · 4 beat it on benchmarks
What papers say
Verbatim critique sentences, each from a paper that cites TrustRAG as a baseline.
Existing defenses treat poisoning as a content-quality problem: filtering bad documents ragdefender, detecting anomalous signals revprag,avfilter, scoring trustworthiness trustrag, or isolating passages robustrag. These approaches share an implicit assumption: if the system can identify poisoned evidence, it will naturally avoid acting on it. We show this assumption is incorrect.
“it makes the unrealistic assumption that malicious documents form a separate cluster in the embedding space.”
“For numerical manipulation: the poisoned passage clusters with the legitimate passage (cosine similarity 0.9997 means they are in the same cluster). Stage 1 clustering cannot separate them. Stage 2 LLM self-assessment would need to notice that `$15,500` ≠ `$15,000` across two passages in the context window, but research on LLM numerical reasoning shows this is unreliable, especially when the numbers are embedded in otherwise identical text.”
“Although promising, these approaches have two major limitations: Majority-voting often fails under high poisoning, while heuristic and aggressive filtering may lose relevant content under low poisoning.”
Beaten on benchmarks
Head-to-head results where a newer method reports beating TrustRAG. Values are copied from the source paper's tables — verify against the cited paper.
RAGShield beats TrustRAG
0.0 vs 89.0
Overall Attack Success Rate (%) · [Synthetic corpus, T3-H T6 T-TEMP Overall]
RAGShield: Provenance-Verified Defense-in-Depth Against Knowledge Base Poisoning in Government Retrieval-Augmented Generation SystemsGMTP beats TrustRAG
3.5 vs 11.5
ASR (Attack Success Rate) · [PoisonedRAG attack, Generation phase, NQ dataset]
Safeguarding RAG Pipelines with GMTP: A Gradient-based Masked Token Probability Method for Poisoned Document DetectionBRIDGE_GRPO beats TrustRAG
85.48 vs 66.41
Accuracy · [Qwen 72B / TRD Simu]
After Retrieval, Before Generation: Enhancing the Trustworthiness of Large Language Models in RAGSeconRAG beats TrustRAG
77.5 vs 75.5
PIA_ACC · [Mistral-12B, HotpotQA]
SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG
What to use instead
Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.
- May 26, 2026
- May 19, 2026
- May 1, 2026
- Beyond Factual GroundingBeyond Factual Grounding: The Case for Opinion-Aware Retrieval-Augmented GenerationApr 13, 2026
- RAGShieldRAGShield: Provenance-Verified Defense-in-Depth Against Knowledge Base Poisoning in Government Retrieval-Augmented Generation SystemsApr 1, 2026
- Mar 24, 2026
- Jan 13, 2026
- Oct 10, 2025
- RADARRADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized CollaborationSep 28, 2025
- RAGOriginWho Taught the Lie? Responsibility Attribution for Poisoned Knowledge in Retrieval-Augmented GenerationSep 17, 2025