Keywords
retrieval self-check
distractor density
paired simulation
reproducibility
Abstract
We evaluated retrieval self-check for retrieval-augmented answering under 9% distractor density. A deterministic paired simulation generated 56 cases and preserved a upper-severity quartile. Mean evidence faithfulness changed from 0.544 to 0.590; the paired difference was +0.046 (95% interval +0.044 to +0.049). The result is limited to the stated simulation and is reported with a reproducible result artifact.
References
Sang, Y. (2025). Towards Explainable RAG: Interpreting the Influence of Retrieved Passages on Generation. 2025 4th International Conference on Robotics, Artificial Intelligence and Intelligent Control (RAIIC), 397-400. https://doi.org/10.1109/raiic65850.2025.11170170
Ji, S., Zhang, X., Huang, Y., & Li, D. (2026). CGS-RAG:Community-Aggregated Graph Semantic Retrieval-Augmented Generation. https://doi.org/10.21203/rs.3.rs-9748268/v1
Köhler, A., Stein, M., & Vogt, E. (2026). Adaptive Context Reconstruction for Enhanced Retrieval Augmented Generation . https://doi.org/10.2139/ssrn.6101386
Liu, L. (2026). A Survey of (Deep RAG) Deep Retrieval Augmented Generation and Reasoning in Large Language Models. https://doi.org/10.36227/techrxiv.177272838.89432844/v1
Roy, P., & Roy, D. (2025). Building a Security and Reliability Evaluation Suite for Retrieval-Augmented Generation (RAG) Systems. https://doi.org/10.20944/preprints202510.0418.v2
S, S. B., & N, K. (2026). UAMG-RAG: Adaptive Multi-Granularity Retrieval-Augmented Generation with Uncertainty-Guided Re-Retrieval. https://doi.org/10.21203/rs.3.rs-10020935/v1
Dhanda, K. (2026). MedRAG: Retrieval-Augmented Generation for Medical QA-Comparing Base and RAG-Augmented LLM Performance on Evidence from Peer-Reviewed Clinical Research. https://doi.org/10.2139/ssrn.6608518
Genesis, J. (2025). Retrieval-Augmented Text Generation: Methods, Challenges, and Applications. https://doi.org/10.20944/preprints202504.0443.v1
Aryani, A. (2024). A Brief Introduction to Retrieval Augmented Generation (RAG). https://doi.org/10.59350/ayxyj-h9751
Srivastava, M. (2026). A Privacy-First Architecture for Fully Local Retrieval-Augmented Generation in Secure Document Intelligence. https://doi.org/10.36227/techrxiv.176800894.46972585/v1