DisastRAG: A Multi-Source Disaster Information Integration and Access System Based on Retrieval-Augmented Large Language Models

This feed contains multiple 2026 papers on retrieval-augmented generation (RAG), retrieval interfaces, recommender systems, and synthetic enterprise datasets. Key contributions include: DisastRAG (multi-path RAG for disaster info combining document, structured records, and web fallback); studies showing retrieval improves LLM performance but that stronger models are more sensitive to retrieval noise; Direct Corpus Interaction (DCI) where agents use terminal tools (grep, file reads, scripts) to search raw corpora without indexing; methods for predicting factual confidence and using conformal/ML