Q2D-Web offers large benchmark for testing web search in AI systems
Q2D-Web: A Large-Scale Benchmark for Retrieval in Agentic RAG Systems
Information RetrievalComputation and Language
Summary
Finding the best way for AI systems to search big collections of web documents is hard because existing tests are too small or not realistic. The authors made a huge test set called Q2D-Web with 190 million documents and 70,000 search queries that come from real conversations, rewritten by AI instead of humans. They included three types of answers to check which documents are really helpful. Testing many search methods, they saw how results can change depending on topic and language but not much by how answers are judged. They also showed you can test on a smaller part of the huge collection and still get good comparisons between search methods.
retrievalRAG systemssearch queriesagent reformulationbenchmarkweb corpusrelevance judgmentsrecallevaluationmultilingual
Authors
Maximilian Schall, Sedigheh Eslami, Markus Krimmel, Antoine Chaffin, Louis Milliken, Bo Wang, Denis Bykov
Abstract
Evaluating first-stage retrievers in large-scale production RAG requires a benchmark that pairs a large-scale corpus with a large set of agent-reformulated search queries based on real user queries and their conversation threads, and that labels many relevant documents per query. No existing public benchmark evaluates this setting: large-scale collections typically provide only a small number of evaluation queries, whereas benchmarks with many queries generally contain only millions of documents. Moreover, most benchmarks assess human-written queries, while the first-stage retrievers in agentic RAG pipelines serve machine-written reformulations whose distribution differs from human search behavior. To overcome these evaluation gaps, we introduce Q2D-Web (Query2Doc-Web), a large-scale agentic retrieval benchmark consisting of a 190M-document web corpus and 70k agentic search queries in ten languages, reformulated from real-world user queries in production systems. Q2D-Web provides three sets of fixed relevance judgments: agent citations, production rankings, and a combined set that unions both signals and adds LLM-based judgments of unlabeled pooled documents to reduce false negatives. We benchmark 13 retrievers including lexical, dense, and late-interaction models and find that their relative ordering is largely insensitive to the choice of judgment set, while diverging substantially across topical domains, query languages, and query types. To enable fast evaluation, we also study subcorpus sampling as an approximation to full-corpus evaluations. Retaining a third of the corpus, selected by reciprocal rank fusion over pooled retriever runs, preserves the full-corpus model ranking under the combined judgments while raising absolute Recall@1000 only by 3 to 7 points. The public leaderboard is accessible under: https://huggingface.co/spaces/perplexity-ai/q2d-web-leaderboard