List your company

Hallucination Detection on BoolQ (ROC-AUC)

94.76ROC-AUCLinear probe
49.093660.949372.80584.6607Jun 20, 2026

Evaluation results

MethodLinks
Linear probeModel=Qwen2.5-7B, Generation Stage=Post-generation2026.0694.76
Attention probeModel=Qwen2.5-7B, Generation Stage=Post-generation2026.0694.46
Attention probeModel=Mistral-7B, Generation Stage=Pre-generation2026.0693.21
Attention probeModel=Mistral-7B, Generation Stage=Post-generation2026.0693.18
Linear probeModel=Mistral-7B, Generation Stage=Post-generation2026.0692.24
Attention probeModel=Qwen2.5-3B, Generation Stage=Post-generation2026.0692.18
Attention probeModel=Qwen2.5-3B, Generation Stage=Pre-generation2026.0692.03
Linear probeModel=Qwen2.5-3B, Generation Stage=Post-generation2026.0691.42
Attention probeModel=Llama-2-7B, Generation Stage=Post-generation2026.0690.94
Linear probeModel=Llama-2-7B, Generation Stage=Post-generation2026.0690.66
Attention probeModel=Llama-2-7B, Generation Stage=Pre-generation2026.0689.88
Attention probeModel=Qwen2.5-7B, Generation Stage=Pre-generation2026.0684.23
Linear probeModel=Qwen2.5-7B, Generation Stage=Pre-generation2026.0682.25
Linear probeModel=Mistral-7B, Generation Stage=Pre-generation2026.0678.99
EntropyModel=Qwen2.5-7B, Generation Stage=Post-generation2026.0678.35
Linear probeModel=Qwen2.5-3B, Generation Stage=Pre-generation2026.0669.65
Linear probeModel=Llama-2-7B, Generation Stage=Pre-generation2026.0668.64
EntropyModel=Qwen2.5-3B, Generation Stage=Post-generation2026.0668.58
Self-check, zero-shotModel=Qwen2.5-7B, Generation Stage=Pre-generation2026.0664.19
Self-check, zero-shotModel=Qwen2.5-3B, Generation Stage=Pre-generation2026.0663.36
EntropyModel=Llama-2-7B, Generation Stage=Post-generation2026.0662.63
Self-check, zero-shotModel=Mistral-7B, Generation Stage=Pre-generation2026.0661.11
Self-check, zero-shotModel=Llama-2-7B, Generation Stage=Pre-generation2026.0652.65
Question lengthModel=Mistral-7B, Generation Stage=Pre-generation2026.0652.52
EntropyModel=Mistral-7B, Generation Stage=Post-generation2026.0651.08
Question lengthModel=Qwen2.5-3B, Generation Stage=Pre-generation2026.0651.08
Question lengthModel=Llama-2-7B, Generation Stage=Pre-generation2026.0651
Question lengthModel=Qwen2.5-7B, Generation Stage=Pre-generation2026.0650.85