FusionRAG-Ex: An Interactive RAG Framework with Retrieval and Confidence-Aware Explanations

Description

FusionRAG-Ex is an open-source, explainability-first Retrieval-Augmented Generation (RAG) framework. It combines three retrieval strategies (BM25, Dense/FAISS, Hybrid) with three importance-scoring modes — Perturbation Dissimilarity (PD), Retrieval-Weighted (RW), and Confidence-Retrieval Fusion — to give researchers and practitioners transparent, unit-level explanations of why a RAG pipeline produces any given answer. The framework ships with a fully interactive web UI, a multi-dataset evaluation pipeline, and a simple JSONL-based data format that lets you plug in any custom document corpus and get up and running in minutes. Github repo: chandanasreemala/XAI_RAG