Natural Questions

Natural Questions is a benchmark for broad factual and academic knowledge. The official description says: "Google Natural Questions - open-domain QA using real user queries from Google Search". Below we describe in simple words what benchmarks of this type generally examine.

Natural Questions measures broad factual and academic knowledge. In detail, it checks:

• facts and concepts from many academic fields at once
• harder variants with several plausible answer options
• knowledge that goes beyond the model's training-set surface forms
• resisting the plausible-but-wrong answers deliberately included

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

Official methodology note: Google Natural Questions - open-domain QA using real user queries from Google Search. 5-shot.

What is the Natural Questions benchmark?

Natural Questions is a benchmark for broad factual and academic knowledge. The official description says: 'Google Natural Questions - open-domain QA using real user queries from Google Search'. Below we describe in simple words what benchmarks of this type generally examine.

What does Natural Questions test?

Natural Questions mainly measures broad factual and academic knowledge. Concretely, it checks: facts and concepts from many academic fields at once; harder variants with several plausible answer options; knowledge that goes beyond the model's training-set surface forms.

How is the Natural Questions score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get Natural Questions data and scores?

The official source is https://github.com/google-research-datasets/natural-questions. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/natural-questions/scores) or pull it directly from the Open LLM Radar API with any HTTP client.