List your company

Hallucination detection on SQuAD v2 (test held-out)

0.8577AUCHierarchicalTransformer
0.391260.5123550.633450.754545Apr 7, 2026

Evaluation results

MethodLinks
HierarchicalTransformerProbe Variant=M3, Status=Best Probe2026.040.857770.5766.44
Token F1 Similaritytype=Token-overlap2026.040.422236.5231.16
Cosine Similarity (Max)type=Embedding-based2026.040.41142.6951.46
Softmax Similaritytype=Embedding-based2026.040.410542.5450.68
Mean Similaritytype=Embedding-based2026.040.409242.249