Researchllmbenchmarksgroundinggoogle deepmind
Google DeepMind Releases FACTS Benchmark Measuring Accuracy
9.0
This week Google DeepMind introduced the FACTS Benchmark Suite, which measures how reliably AI models produce factually accurate answers across four tasks: internal factoids, web search use, long-document grounding, and image interpretation. The best model, Gemini 3 Pro, scored 69% accuracy, with others substantially lower, highlighting that models are wrong about roughly one-third of answers and raising sectoral risks.
Key Points
- 1Introduces FACTS benchmark testing models on internal facts, web search, long documents, and images
- 2Shows limited factual reliability: Gemini 3 Pro hits 69% accuracy, other leading models score much lower
- 3Warns practitioners: factual errors risk legal, financial, and healthcare harms; require verification workflows
Scoring Rationale
Major official benchmark provides industry-wide factuality measurements, but it's primarily diagnostic rather than delivering immediate technical fixes.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
High-Value Overnight OrdersEasyDelivered International ShipmentsMediumOn-Time Delivery Rate by CarrierHard
250 free problems · No credit card
See all Logistics & Shipping problems