List your company
ResearchFormerly WizWand

Hallucination Detection on PubmedQA

88F1 ScoreLatentAudit
4.48826.16947.8569.531Nov 12, 2025Dec 6, 2025Dec 30, 2025Jan 24, 2026Feb 17, 2026Mar 13, 2026Apr 7, 2026
Updated 6mo ago

Evaluation results

MethodLinks
LatentAuditBase Model=Qwen-32026.0488—94.8
GPT-4o JudgeBase Model=Llama-22026.0487—94.8
GPT-4o JudgeBase Model=Llama-32026.0487—94.8
GPT-4o JudgeBase Model=Qwen-2.52026.0487—94.8
GPT-4o JudgeBase Model=Qwen-32026.0487—94.8
GPT-4o JudgeBase Model=Mistral2026.0487—94.8
LatentAuditBase Model=Llama-32026.0487—94.2
LatentAuditBase Model=Llama-22026.0486—93.1
LatentAuditBase Model=Qwen-2.52026.0486—93.8
LatentAuditBase Model=Mistral2026.0485—92.5
INSIDEBase Model=Llama-32026.0484—90.8
INSIDEBase Model=Llama-22026.0483—90.3
INSIDEBase Model=Qwen-32026.0483—90.5
INSIDEBase Model=Qwen-2.52026.0482—89.9
INSIDEBase Model=Mistral2026.0482—89.5
SAPLMABase Model=Llama-32026.0482—88.2
HalluCleanBackbone=GPT-3.5-turbo2025.1181.781—
SelfCheckGPTBase Model=Llama-32026.0481—87.1
SAPLMABase Model=Llama-22026.0481—87.8
SAPLMABase Model=Qwen-32026.0481—88
SelfCheckGPTBase Model=Llama-22026.0480—86.2
SelfCheckGPTBase Model=Qwen-32026.0480—86.9
SAPLMABase Model=Qwen-2.52026.0480—87.2
SAPLMABase Model=Mistral2026.0480—87
SelfCheckGPTBase Model=Qwen-2.52026.0479—85.5
SelfCheckGPTBase Model=Mistral2026.0479—85.8
DeepSeek-R12025.1176.479—
DeepSeek-V32025.117177.5—
Min-PerplexityBase Model=Llama-32026.0470—72.2
Min-PerplexityBase Model=Llama-22026.0469—71.8
Min-PerplexityBase Model=Qwen-32026.0469—72
GPT-4o-mini2025.1168.774.5—
Min-PerplexityBase Model=Qwen-2.52026.0468—71.5
Min-PerplexityBase Model=Mistral2026.0468—71.2
Llama-3-70B2025.1116.554.5—
GPT-3.5-turbo2025.117.752—