List your company

Token-level Hallucination Detection on AIME 2025 (AUROC/AUPRC)

89.47AUROCTOKENHD-8B
63.345270.127676.9183.6924May 12, 2026

Evaluation results

MethodLinks
TOKENHD-8B2026.0589.4780.14
o4-mini2026.0582.2660.88
GPT-4.12026.0575.4349.87
QwQ-32B2026.0570.8252.12
R1-Qwen3-8B2026.0564.3546.63