AI-generated text detection on HC3 (test) benchmark leaderboard · SOTA2 Research

AI-generated text detection on HC3 (test)

99.93F1 (Overall)SCRN

Evaluation Results

Method	Links
SCRN2024.06		99.93	99.78	100	100	99.9	—
Deberta2024.06		99.92	99.74	100	100	99.88	—
Flooding2024.06		99.89	99.67	100	100	99.85	—
RDrop2024.06		99.88	99.67	99.96	99.98	99.85	—
Roberta2024.06		99.8	99.45	100	100	99.74	—
ChatGPT-Detector2024.06		99.58	99.03	99.15	99.61	99.56	—
Bert2024.06		99.45	98.28	100	100	99.2	—
SeqXGPT2024.06		99.33	97.92	100	100	99.03	—
PECOLA2024.02		99.15	—	—	—	—	99.15
PECOLA2024.02		98.63	—	—	—	—	98.63
PECOLA2024.02		98.59	—	—	—	—	98.59
RMLM2024.06		98	96.15	100	100	96	—
GLTR2024.06		97.68	96.84	95.76	98.06	98.57	—
Log-Rank2024.06		97.54	96.3	95.87	98.11	98.31	—
PECOLA2024.02		97.19	—	—	—	—	97.19
Log-Likelihood2024.06		97.18	95.43	95.61	97.98	97.9	—
RanMASK2024.06		95.67	87.73	100	100	93.58	—
Entropy2024.06		93	89.9	87.69	94.41	95.47	—