Google FACTS Grounding scores whether long-form answers stay faithful to a supplied document, using an ensemble of LLM judges plus an eligibility filter.
unassessed
| Category | generation |
|---|---|
| Subcategory | long-form answers grounded in a supplied document (up to 32k tokens) |
| Page status | active |
| Metric | final factuality score (share of responses that are eligible and fully grounded, averaged over three judges) |
| Direction | higher_is_better |
| Unit | % |
| Dataset size | 1719 |
| Dataset licence | CC-BY-4.0 |
| Publisher | Google DeepMind, Google Research, Google Cloud, and Kaggle |
FACTS Grounding checks scenario-one factuality: every claim in a long-form answer must be supported by a document in the prompt, not by parametric world knowledge. Each example has a system instruction ("use only this context"), a user request (QA, summary, rewrite) and a web-sourced document up to 32k tokens (mean about 2.5k). The request must be non-trivial and must not need extra-document expertise, maths, or creative writing. This is attribution to provided context, not closed-book factoids and not web search.
Single-turn generation: system instruction plus user request plus full document. Judges then run two stages, eligibility then span/response-level grounding.
No model card in ModelSpec reports this benchmark yet.