Skip to content

Benchmark

OmniScience · Non-Hallucination

Factuality benchmarks measure how often models get facts right. Scores are percentages. Higher is better.

Models measured
1
Scores recorded
1
Sources
1
IQ weight
No weight published

Score distribution

78.7%
Mean
78.7%
Median
0%
Std dev
78.7%
10th percentile
78.7%
90th percentile
78.7%
Min

Leaderboard

#ModelScore
1Nemotron 3 UltraNeMo Evaluator SDK · official table· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.78.7%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.