Updated daily at 05:00 UTC
Benchmark
FACTS Benchmark Suite
Measures: The FACTS Score is the average accuracy over four sub-benchmarks, each averaged over public and private sets. Parametric is closed-book factoids, judged by Gemini 2.5 Pro. Search is answering with a web-search tool, judged by Gemini 2.0 Flash. Multimodal is image questions graded against human-written rubrics by an automated judge that is not named. Grounding v2 is long-document answers, judged by Gemini 2.5 Flash and GPT-5.
- Maintainer
- Google DeepMind / Google Research; Kaggle runs the leaderboard and holds the private sets
- Introduced
- 2025
- Status
- Active
- Saturation
- Headroom
- Typical provenance
- Independent
- In scoring set
- Yes
Leaderboard
Independent and third-party results only.
No independent or third-party results yet. Show all results