List your company
ResearchFormerly WizWand

Hallucination Detection on TyDiQA-GP

0.9404AUC ROCHaloScope
0.4733360.5945930.715850.837107Dec 8, 2025Jan 5, 2026Feb 3, 2026Mar 4, 2026Apr 1, 2026Apr 30, 2026May 29, 2026
Updated 5mo ago

Evaluation results

MethodLinks
HaloScopeModel=LLaMA-2-7B2025.120.9404
HALLUSHIFT++Model=LLaMA-2-7B2025.120.8966
HALLUSHIFTModel=LLaMA-2-7B2025.120.8761
HALLUSHIFT++Model=OPT-6.7B2025.120.8758
HALLUSHIFTModel=OPT-6.7B2025.120.8511
HaloScopeModel=OPT-6.7B2025.120.8098
CCS*Model=LLaMA-2-7B2025.120.8038
EigenScoreBackbone=LLaMA-3-8B-Instruct2026.050.7463
FLaG (Full)Backbone=LLaMA-3-8B-Instruct2026.050.7441
Self-evaluationBackbone=LLaMA-3-8B-Instruct2026.050.7427
HaloScopeBackbone=LLaMA-3-8B-Instruct2026.050.7408
EarlyDetecBackbone=LLaMA-3-8B-Instruct2026.050.7288
FLaG (Semi)Backbone=LLaMA-3-8B-Instruct2026.050.7227
FLaG (Full)Backbone=Qwen2.5-7B-Instruct2026.050.7205
CCSBackbone=LLaMA-3-8B-Instruct2026.050.7202
Linear probeBackbone=Qwen2.5-7B-Instruct2026.050.7146
EGHBackbone=LLaMA-3-8B-Instruct2026.050.7124
TTPDBackbone=Qwen2.5-7B-Instruct2026.050.7063
Linear probeBackbone=LLaMA-3-8B-Instruct2026.050.7062
TSVBackbone=LLaMA-3-8B-Instruct2026.050.7015
EarlyDetecBackbone=Qwen2.5-7B-Instruct2026.050.6995
TTPDBackbone=LLaMA-3-8B-Instruct2026.050.6919
SAPLMABackbone=LLaMA-3-8B-Instruct2026.050.6903
SPUQBackbone=LLaMA-3-8B-Instruct2026.050.674
Probe-LRBackbone=LLaMA-3-8B-Instruct2026.050.6708
FLaG (Semi)Backbone=Qwen2.5-7B-Instruct2026.050.6702
SAPLMABackbone=Qwen2.5-7B-Instruct2026.050.6674
HaloScopeBackbone=Qwen2.5-7B-Instruct2026.050.6622
Probe-LRBackbone=Qwen2.5-7B-Instruct2026.050.662
TSVBackbone=Qwen2.5-7B-Instruct2026.050.6583
CCS*Model=OPT-6.7B2025.120.6462
EGHBackbone=Qwen2.5-7B-Instruct2026.050.6417
Lexical SimilarityBackbone=LLaMA-3-8B-Instruct2026.050.6266
EigenScoreBackbone=Qwen2.5-7B-Instruct2026.050.6248
Lexical SimilarityBackbone=Qwen2.5-7B-Instruct2026.050.6198
SPUQBackbone=Qwen2.5-7B-Instruct2026.050.6124
Self-evaluationBackbone=Qwen2.5-7B-Instruct2026.050.5759
SelfCKGPTBackbone=Qwen2.5-7B-Instruct2026.050.5601
Semantic EntropyBackbone=LLaMA-3-8B-Instruct2026.050.5569
CCSBackbone=Qwen2.5-7B-Instruct2026.050.5544
VerbalizeBackbone=LLaMA-3-8B-Instruct2026.050.5486
VerbalizeBackbone=Qwen2.5-7B-Instruct2026.050.5365
SelfCKGPTBackbone=LLaMA-3-8B-Instruct2026.050.5283
PerplexityBackbone=LLaMA-3-8B-Instruct2026.050.5225
Semantic EntropyBackbone=Qwen2.5-7B-Instruct2026.050.5021
PerplexityBackbone=Qwen2.5-7B-Instruct2026.050.4913