←Back to NewsAI News/RetrievalpaperRetrievalBenchmarksSTAIR Beats Every RAG Baseline by Reading Documents Like a Table of ContentsA new generative retriever leverages the table of contents to hit 82.6% Recall@1 on a fresh 18-book benchmark, with hallucinations under 0.05%.SourceAlphaSignalPublishedSep 3, 2026, 2:02 PMAuthorAlphaSignal NewsroomRead1 min readA new generative retriever leverages the table of contents to hit 82.6% Recall@1 on a fresh 18-book benchmark, with hallucinations under 0.05%.Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.Read original report ↗Next readsPerplexity · newsPerplexity's Q2D-Web Benchmark Tests AI Search on 190M Real Web DocumentsKaggle · newsLlamaIndex's ExtractBench Catches GPT-5 and Gemini-3 Hitting 0% on Wide SchemasArtificial Analysis · newsPerplexity's Search API Tops Agent Leaderboard, Beating Rivals at $5 Flat