ResearchBenchmarksHuman Evaluation on DeepResearch Bench 20 reports (sampled)Follow95Readability (Win/Tie Rate)PTAH84.687.39092.7May 28, 2026Evaluation ResultsMethodMethodLinksReadability (Win/Tie Rate)Usability (Win/Tie Rate)Information Acquisition Efficiency (Win/Tie Rate)Overall Preference (Win/Tie Rate)PTAHEvaluator=General U2Evaluator=General U22026.0595909595PTAHEvaluator=General U1Evaluator=General U12026.059095100100PTAHEvaluator=AverageEvaluator=Average2026.0588.7588.7596.2595PTAHEvaluator=Expert E1Evaluator=Expert E12026.0585909595PTAHEvaluator=Expert E2Evaluator=Expert E22026.0585809590