mlsanityData to 14 Oct 2026

MIRAGE: Defending Long-Form RAG Against Misinformation Pollution

arXiv 2607.05069 ↗First posted 6 Jul 2026cs.CLcited by 0

On the map: Retrieval-augmented generation → RAG robustness and adaptive retrieval

Abstract (arXiv)

Retrieval-Augmented Generation (RAG) improves factuality by grounding LLMs in external evidence, but real-world retrieval is often polluted: semantically relevant passages may contain subtle misinformation, misleading framings, or fabrications. We introduce MIRAGE, a training-free, model-agnostic defense for long-form RAG. MIRAGE builds an NLI-based cross-document claim graph and applies a Defended-Claims Gate to either condition generation on a consistent, multi-source supported subset or to block retrieval and answer parametrically. We also release a minimal-edit pollution protocol spanning four perturbation families (Unambiguous, Conflicting, Misleading, Fabricated) to construct matched clean, mixed, and fully polluted evaluation regimes. Across four long-form QA benchmarks and multiple commercial and open-weight LLMs, pollution severely degrades vanilla RAG, while MIRAGE consistently restores factuality under mixed and fully polluted evidence and outperforms prior robust-RAG methods. Our implementation and datasets are available at https://github.com/SaadElDine/MIRAGE.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation. Builds on 15 of its 29 references that are in the map. First posted on arXiv 6 Jul 2026.

Builds on

15 of 29 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025After Retrieval, Before Generation: Enhancing the Trustworthiness of Large Language Models in Retrieval-Augmented Generation21 May
  2. 2025Retrieval-Augmented Generation with Conflicting Evidence17 Apr
  3. 2024InstructRAG: Instructing Retrieval-Augmented Generation via Self-Synthesized Rationales19 Jun
  4. 2024Evidence-Driven Retrieval Augmented Response Generation for Online Misinformation22 Mar
  5. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection17 Oct

Other references · 10

  1. 2020DeBERTa: Decoding-enhanced BERT with Disentangled Attentioncited by 2,181
  2. 2023FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generationcited by 1,282
  3. 2023AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedbackcited by 748
  4. 2023Making Retrieval-Augmented Language Models Robust to Irrelevant Contextcited by 369
  5. 2024Certifiably Robust RAG against Retrieval Corruptioncited by 151
  6. 2024Fine-grained Hallucination Detection and Editing for Language Modelscited by 149
  7. 2024VERISCORE: Evaluating the factuality of verifiable claims in long-form text generationcited by 117
  8. 2024Improving Factuality with Explicit Working Memorycited by 10
  9. 2025VeriFastScore: Speeding up long-form factuality evaluationcited by 8
  10. 2025FaStfact: Faster, Stronger Long-Form Factuality Evaluations in LLMscited by 8

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Worse than Zero-shot? A Fact-Checking Dataset for Evaluating the Robustness of RAG Against Misleading Retrievals
  2. 2026Evaluating RAG Reliability under Clean, Misleading, and Mixed Retrieval
  3. 2025FAIR-RAG: Faithful Adaptive Iterative Refinement for Retrieval-Augmented Generation
  4. 2025MIRAGE: A Metric-Intensive Benchmark for Retrieval-Augmented Generation Evaluation
  5. 2026Towards Query-Agnostic RAG Evaluation via Query Coverage and Claim Verifiability
  6. 2026RAGStress: A controlled benchmark for evaluating retrieval-augmented generation under knowledge-base degradation
  7. 2025Resolving Conflicting Evidence in Automated Fact-Checking: A Study on Retrieval-Augmented LLMs
  8. 2025Rethinking All Evidence: Enhancing Trustworthy Retrieval-Augmented Generation via Conflict-Driven Summarization