List your company

Hallucination Detection on CoQA (AUROC, F1)

0.98AUROCPCNET
0.3040.47950.6550.8305May 7, 2026

Evaluation results

MethodLinks
PCNETModel=Llama-3.1-8B2026.050.980.98
PCNETModel=Mistral-7B-v0.32026.050.970.96
HaloScopeModel=Llama-3.1-8B2026.050.970.97
PCNETModel=Qwen3-4B2026.050.960.96
PCNETModel=Llama-3.2-1B2026.050.950.95
HaloScopeModel=Mistral-7B-v0.32026.050.880.88
HaloScopeModel=Llama-3.2-1B2026.050.850.85
SEPModel=Llama-3.2-1B2026.050.770.77
SEPModel=Mistral-7B-v0.32026.050.690.72
SEPModel=Qwen3-4B2026.050.650.73
HaloScopeModel=Qwen3-4B2026.050.630.73
AutoFactModel=Llama-3.2-1B2026.050.560.67
Token NLLModel=Mistral-7B-v0.32026.050.560.67
AutoFactModel=Mistral-7B-v0.32026.050.560.67
AutoFactModel=Qwen3-4B2026.050.560.67
AutoFactModel=Llama-3.1-8B2026.050.560.67
Token NLLModel=Llama-3.2-1B2026.050.550.67
Token NLLModel=Qwen3-4B2026.050.550.67
Token NLLModel=Llama-3.1-8B2026.050.550.67
SEPModel=Llama-3.1-8B2026.050.330.67