DOTRAG: Retrieval-Time Reasoning Along Paths

This feed contains multiple papers on Retrieval-Augmented Generation (RAG) and GraphRAG advances and one security-focused paper (ALDEN) that materially raises privacy risks. ALDEN demonstrates a practical, high‑efficiency private data extraction attack against RAG systems by combining active learning to diversify malicious queries with a decay-based dynamic algorithm to estimate topic/distribution of the underlying knowledge base — substantially outperforming prior extraction methods. Related papers (DotRAG, STAR, ClusterRAG, Agentic GraphRAG, ClinQueryAgent, WAG, etc.) improve query‑condition