Safety IndexEvaluationsFACTS Grounding
FACTS Grounding
Interaction and honesty · Factuality and hallucinationCounts in the indexRun by SAI
How often the model's long-form answers stay fully grounded in the document it was given.
- Metric
- % grounded. Higher is better.
- Dataset
- www.kaggle.com/facts-leaderboard · 860 public · CC-BY-4.0
- Grader
- two-model panel (GPT-4.1 + Gemini 2.5 Flash) approximating Google's panel (the paper used 3-frontier-judge panel (v2 upgraded))
- Status
- Run by SAI. Measured on 3 models, 3 readings by SAI.
- Replicability
- partial
Results
| Model | Value | Score | Source | Setting |
|---|---|---|---|---|
| 92.3%n = 860 | 92 | SAI Measured by SAI · 2026-10-23 | Run by SAI through the model's public API with the provider's default sampling and reasoning left on.Full configurationapproximate 2-judge panel gpt-4.1 + google/gemini-2.5-flash (not Google's panel); released v1 json templates; 860 public examples; provider-default sampling, reasoning on | |
| 89.8%n = 860 | 90 | SAI Measured by SAI · 2026-10-22 | Run by SAI through the model's public API with the provider's default sampling and reasoning left on.Full configurationapproximate 2-judge panel gpt-4.1 + google/gemini-2.5-flash (not Google's panel); released v1 json templates; 860 public examples; provider-default sampling, reasoning on | |
| 84.4%n = 860 | 84 | SAI Measured by SAI · 2026-10-22 | Run by SAI through the model's public API with the provider's default sampling and reasoning left on.Full configurationapproximate 2-judge panel gpt-4.1 + google/gemini-2.5-flash (not Google's panel); released v1 json templates; 860 public examples; provider-default sampling, reasoning on |