Benchmarks

February 24, 2026 ยท View on GitHub

Standardized evaluation suites for measuring research quality.

BenchmarkWhat it measuresDataset sizeAgents tested
FreshQAFactual accuracy on current knowledge600 questionsShallow, Full pipeline
Deep Research BenchReport quality (RACE + FACT metrics)100 topicsDeep researcher
DeepSearchQADocument QA across categories900 problemsDeep researcher
:titlesonly:

freshqa.md
deep-research-bench.md
deepsearch-qa.md