HC3
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
HC3 PWWS attack, AI to Human (in-domain)
100Overall Accuracy
28
HC3 AI-Prob
84.1Finance Accuracy
24
HC3 (test)
99.93F1 (Overall)
18
HC3 Pruthi attack Overall (in-domain)
100Overall Accuracy
14
HC3 Pruthi attack Human to AI (in-domain)
1OA
14
HC3 Pruthi attack AI to Human (in-domain)
100Overall Accuracy
14
HC3 Deep-Word-Bug attack Overall (in-domain)
100OA
14
HC3 Deep-Word-Bug attack Human to AI (in-domain)
100Overall Accuracy
14
HC3 Deep-Word-Bug attack AI to Human (in-domain)
100Overall Accuracy
14
HC3 PWWS attack, Human to AI (in-domain)
100OA
14
HC3 COLING2025 (val)
99.99AUC
13
HC3 Gemini-3.1 pro preview generated text
0.813ROC AUC
10
HC3 Gemini-3 flash preview generated text
85.3ROC AUC
10
HC3 Gemini-3.1 flash-lite preview generated text
0.86ROC AUC
10
HC3 GPT-5.4 mini generated text
0.967ROC AUC
10
HC3 GPT-5.4 generated text
0.974ROC AUC
10
HC3
100AUROC
10
HC3 (full)
99.92Accuracy
9
HC3 Plus (test)
0.985AUROC
5
HC3 (test)
0.995AUROC
5
HC3 cross-benchmark transfer
99.8AUROC (Finance)
3
HC3 Chi-Psy
90TPR @ FPR=1%
3
HC3 Cross-domain
80.48Score (Medicine)
3
HC3 PLUS (test_si)
87.21Balanced Accuracy (BA)
2
HC3 PLUS QA (test)
99.47Balanced Accuracy
2