ResearchBenchmarksTask Execution on TEXTCRAFT-SYNTH 8K context Hard (evaluation set)Follow88Success RateRecursive-3.5220.244467.76May 7, 2026Evaluation ResultsMethodMethodLinksSuccess RateSteps TakenTime (s)RecursiveContext Window=8K trai...Context Window=8K train, 8K eval2026.0588——SingleContext Window=8K trai...Context Window=8K train, 8K eval2026.050——