GlossaryEvaluation
FRAMES benchmark
A retrieval evaluation benchmark testing multi-hop, freshness-sensitive web QA.
FRAMES (Factuality, Retrieval, And reasoning Measurement Set, Krishna et al. 2024, arXiv:2409.12941) is a benchmark for evaluating end-to-end RAG systems. It contains queries that require multi-hop reasoning over retrieved web content. It has become a standard yardstick for comparing web retrieval providers because it isolates retrieval quality from LLM quality.