Benchmarks

Standardized evaluation suites for measuring research quality.

Benchmark	What it measures	Dataset size	Agents tested
FreshQA	Factual accuracy on current knowledge	600 questions	Shallow, Full pipeline
Deep Research Bench	Report quality (RACE + FACT metrics)	100 topics	Deep researcher
DeepSearchQA	Document QA across categories	900 problems	Deep researcher

• FreshQA Evaluator
• Deep Research Bench Evaluation of NVIDIA AI-Q Blueprint
• DeepSearchQA Evaluation for AI-Q Deep Researcher