Q-CARE: A Reference-Free RAG Evaluation Framework Accepted to COLM 2026

_reachsumit · x · 2026-08-13

The paper introduces Q-CARE, a reference-free framework for evaluating Retrieval-Augmented Generation (RAG). By decomposing queries into sub-queries and answers into atomic claims, it enables fine-grained assessment based on query coverage and claim verifiability.

On a human-annotated benchmark spanning eight datasets, Q-CARE achieves higher correlation with human judgments than existing metrics like RAGEval and RAGChecker, proving its effectiveness as a reliable, automated evaluation tool.

Original post →

More from Research

Research channel →