Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

DeepResearchBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Research TaskDeepResearchBench (DRB)
Accuracy (%)95.9
21
Search-based Question AnsweringDeepResearchBench (test)
Component Score34.63
12
Open-ended writingDeepresearchBench
Overall Score46.93
11
Agentic WritingDeepResearchBench
Pass@149.6
5
Deep Research EvaluationDeepResearchBench
WQ63.95
3
Showing 5 of 5 rows