FACTS Grounding
Factual grounding%Factual grounding evaluation — measures whether a model's answer is supported by the provided source documents.
Updated 4 days agoLatest measured Apr 18, 20265 verified · 0 self-reported
Verified results come from third-party or public leaderboard sources. Self-reported results come from provider papers, blogs, or vendor disclosures and should be compared with extra caution.
At a glance
Total results
5
Models tested
5
Providers
2
Verified · Self-reported
5 · 0
Average
68.13 %
Median
63.76 %
Range
30.81 – 91.86 %
Latest result
Apr 18, 2026
Score distribution
1
0
0
0
0
2
0
0
Methodology
Source-attributed QA where each answer must cite the provided document. Scored on precision and factual accuracy.
Limitations
English-only. Doesn't test recall when the fact is absent.
By provider
- Average: 71.08 %Best: 91.86 %
- Anthropic· 2 models63.76 %Claude Haiku 4.5Average: 63.7 %Best: 63.76 %
Full leaderboard
Showing 5 of 5| # | Model | Provider | Score (%) |
|---|---|---|---|
| 1 | GPT-5.4 | OpenAI | 91.86 |
| 2 | GPT-5.4 nano | OpenAI | 90.58 |
| 3 | Claude Haiku 4.5 | Anthropic | 63.76 |
| 4 | Claude Sonnet 4.6 | Anthropic | 63.64 |
| 5 | GPT-5.4 mini | OpenAI | 30.81 |
Community ratings
No ratings yet. Be the first to rate FACTS Grounding.
Rate FACTS Grounding
Sign in to rate and review.
Comments
Sign in to leave a comment.