Skip to content

Benchmark

LLM-AggreFact MeetB

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

74.8%
Mean
77.3%
Median
7.8%
Std dev
65.8%
10th percentile
81.2%
90th percentile
50.2%
Min
50.2%84.6%
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, MeetB, renderer-v1· LLM-AggreFact data72.6%
Bespoke-Minicheck-7Bgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data78.3%
Claude-2.1grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data72.3%
Claude-3 Opusgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data82.4%
Claude-3.5 Sonnetgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data84.6%
Clefgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data81.1%
Clef Flashgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data80.5%
d1-ultragrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data79.9%
DAEgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data69.5%
Decider v1.2grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data80.3%
Drex v1.5grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data75.6%
FactCG-DeBERTa-Lgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data74.3%
Gemini-Progrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data65.8%
GPT-3.5-Turbogrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data73.4%
gpt-4-turbo-previewgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data79.9%
gpt-4o-2024-05-13grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data79.8%
gpt-4o-mini-2024-07-18grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data79.7%
Granite Guardian 3.3grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data78.6%
HHEM-2.1-opengrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data71%
InternLM2.5-7B-chatgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data68%
Jev 1.13.0grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data83.7%
Llama 3.2 1B InstructLlama-3.2-1B-Instruct, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data50.2%
Llama 3.2 3BLlama-3.2-3B-Instruct, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data52.5%
Llama 3.3 70B InstructLlama-3.3-70B-Instruct, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data78.4%
Llama-3-70B-Instructgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data80.6%
Llama-3.1-405B-Instructgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data81.2%
Llama-3.1-70B-Instructgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data81%
Llama-3.1-8B-Instructgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data75.5%
MiniCheck-DeBERTa-Lgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data72.7%
MiniCheck-Flan-T5-Lgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data77.3%
MiniCheck-RoBERTa-Lgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data75.9%
Mistral-8x7Bgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data73.3%
Mistral-Largegrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data78.9%
Mistral-Large 2grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data84.2%
Mixtral-8x22Bgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data78.5%
nimble-v4grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data79.7%
nimble-v4-maxgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data82.4%
PaLM2-Bisongrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data73.6%
Qwen2.5 0.5b InstructQwen2.5-0.5B-Instruct, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data64.7%
Qwen2.5 72B InstructQwen2.5-72B-Instruct, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data80.4%
Qwen2.5 7B InstructQwen2.5-7B-Instruct, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data75.2%
QwQ 32B PreviewQwQ-32B-Preview, grounded claim checker, MeetB, renderer-v1· LLM-AggreFact data78.5%
SummaC-CVgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data62.8%
SummaC-ZSgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data71%
T5-NLI-Mixedgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data55.3%
Tülu-3-70Bgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data70.6%
Tülu-3-8Bgrounded claim checker, MeetB, renderer-v1· LLM-AggreFact data67.7%

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