Skip to content

Benchmark

LLM-AggreFact all-11 mean

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

Models and systems measured
47
Scores recorded
47
Sources
1
IQ weight
No weight published

Score distribution

71.7%
Mean
73.7%
Median
6.2%
Std dev
63.6%
10th percentile
77.8%
90th percentile
50.3%
Min
50.3%79.5%
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
AlignScoregrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data70.5%
Bespoke-Minicheck-7Bgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data77.4%
Claude-2.1grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data71.1%
Claude-3 Opusgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data74.8%
Claude-3.5 Sonnetgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data77.2%
Clefgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data79%
Clef Flashgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data76.6%
d1-ultragrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data78.5%
DAEgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data65%
Decider v1.2grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data78.7%
Drex v1.5grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data76.5%
FactCG-DeBERTa-Lgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data75.6%
Gemini-Progrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data65.4%
GPT-3.5-Turbogrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data70.1%
gpt-4-turbo-previewgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data76.2%
gpt-4o-2024-05-13grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data75.9%
gpt-4o-mini-2024-07-18grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data74%
Granite Guardian 3.3grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data76.5%
HHEM-2.1-opengrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data71.8%
InternLM2.5-7B-chatgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data63.6%
Jev 1.13.0grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data78.7%
Llama 3.2 1B InstructLlama-3.2-1B-Instruct, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data50.3%
Llama 3.2 3BLlama-3.2-3B-Instruct, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data60%
Llama 3.3 70B InstructLlama-3.3-70B-Instruct, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data74.5%
Llama-3-70B-Instructgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data73.7%
Llama-3.1-405B-Instructgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data74.4%
Llama-3.1-70B-Instructgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data75.1%
Llama-3.1-8B-Instructgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data70.3%
MiniCheck-DeBERTa-Lgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data73.1%
MiniCheck-Flan-T5-Lgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data75%
MiniCheck-RoBERTa-Lgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data73.5%
Mistral-8x7Bgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data67.4%
Mistral-Largegrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data74.2%
Mistral-Large 2grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data76.5%
Mixtral-8x22Bgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data71.5%
nimble-v4grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data77.8%
nimble-v4-maxgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data79.5%
PaLM2-Bisongrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data66.1%
Qwen2.5 0.5b InstructQwen2.5-0.5B-Instruct, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data64.2%
Qwen2.5 72B InstructQwen2.5-72B-Instruct, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data75.6%
Qwen2.5 7B InstructQwen2.5-7B-Instruct, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data72.8%
QwQ 32B PreviewQwQ-32B-Preview, grounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data71.8%
SummaC-CVgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data61%
SummaC-ZSgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data67.7%
T5-NLI-Mixedgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data61.1%
Tülu-3-70Bgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data67.6%
Tülu-3-8Bgrounded claim checker, all-11 mean, renderer-v1· LLM-AggreFact data63.2%

Data credits: LLM-AggreFact, Tang et al. Public numeric grounded checker balanced accuracy; results licence not separately established.