FACTS Parametric
Google DeepMind / Kaggle / Knowledge · Answers factual questions without search to assess the model's own knowledge accuracy.
What it measures and how
Only Score for a fixed task version is considered. Public/private sets and intervals are retained for review, without counting three related scores separately. FACTS is not among AA v4.3's 10 evaluations.
How this evidence is used
Under review: protocols and licenses are confirmed, but completion counts or flags are missing. Some intervals are unusually wide; total/public/private aggregation needs clarification. Awaiting evaluator explanation instead of removing anomalies to force inclusion. No scoring budget is assigned.
Evaluation limits & attribution
Google supplies judges and the evaluation, requiring independent corroboration. Exports do not establish whether some rows completed the same full set. Partial results cannot represent full testing; interval-inferred sample sizes are not officially confirmed facts.
Data license: Apache 2.0 · Explicitly declared on official benchmark and Score task pages. Kaggle documentation provides public JSON downloads without login.
Results are published by Google DeepMind / Kaggle. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.