AI-generated text detection on XSum Generated by GPT-4 (test)
0.9996AUROCDPIC
Evaluation Results
| Method | Links | |
|---|---|---|
| DPICAuxiliary Model=ChatGPT2023.05 | 0.9996 | |
| DPIC(ChatGPT)Interference=Original2023.05 | 0.9996 | |
| DPIC(ChatGPT)Interference=DIPPER2023.05 | 0.9991 | |
| DPICAuxiliary Model=Vicuna-7b2023.05 | 0.9986 | |
| DPIC(Vicuna-7b)Interference=Original2023.05 | 0.9986 | |
| DPIC(Vicuna-7b)Interference=DIPPER2023.05 | 0.9969 | |
| RADARStrategy=Supervised2023.05 | 0.9931 | |
| RADARInterference=Original2023.05 | 0.9931 | |
| DPIC(ChatGPT)Interference=Back-translate2023.05 | 0.9931 | |
| RADARInterference=DIPPER2023.05 | 0.9924 | |
| DPIC(Vicuna-7b)Interference=Back-translate2023.05 | 0.9903 | |
| RADARInterference=Back-translate2023.05 | 0.9608 | |
| Fast-DetectGPTInterference=Back-translate2023.05 | 0.9137 | |
| Fast-DetectGPTStrategy=Zero-shot2023.05 | 0.9064 | |
| Fast-DetectGPTInterference=Original2023.05 | 0.9064 | |
| LikelihoodInterference=Back-translate2023.05 | 0.8449 | |
| LogRankInterference=Back-translate2023.05 | 0.8436 | |
| Fast-DetectGPTInterference=DIPPER2023.05 | 0.8057 | |
| LRRInterference=Back-translate2023.05 | 0.8003 | |
| LikelihoodStrategy=Zero-shot2023.05 | 0.7982 | |
| LikelihoodInterference=Original2023.05 | 0.7982 | |
| LogRankStrategy=Zero-shot2023.05 | 0.798 | |
| LogRankInterference=Original2023.05 | 0.798 | |
| DNA-GPT(Neo-2.7)Interference=Back-translate2023.05 | 0.7776 | |
| ROBERTa-baseInterference=Back-translate2023.05 | 0.7536 | |
| LRRStrategy=Zero-shot2023.05 | 0.7453 | |
| LRRInterference=Original2023.05 | 0.7453 | |
| DNA-GPTAuxiliary Model=Neo-2.72023.05 | 0.7267 | |
| DNA-GPT(Neo-2.7)Interference=Original2023.05 | 0.7267 | |
| RoBERTa-largeBackbone=RoBERTa-large2023.05 | 0.6879 | |
| ROBERTa-largeInterference=Original2023.05 | 0.6879 | |
| ROBERTa-largeInterference=DIPPER2023.05 | 0.6833 | |
| NPRInterference=Back-translate2023.05 | 0.682 | |
| RoBERTa-baseBackbone=RoBERTa-base2023.05 | 0.6778 | |
| ROBERTa-baseInterference=Original2023.05 | 0.6778 | |
| ROBERTa-largeInterference=Back-translate2023.05 | 0.666 | |
| LRRInterference=DIPPER2023.05 | 0.6607 | |
| ROBERTa-baseInterference=DIPPER2023.05 | 0.6469 | |
| LogRankInterference=DIPPER2023.05 | 0.6432 | |
| LikelihoodInterference=DIPPER2023.05 | 0.6296 | |
| DNA-GPTAuxiliary Model=ChatGPT2023.05 | 0.6146 | |
| DNA-GPT(ChatGPT)Interference=Original2023.05 | 0.6146 | |
| DNA-GPT(ChatGPT)Interference=Back-translate2023.05 | 0.607 | |
| DNA-GPT(Vicuna-7b)Interference=Back-translate2023.05 | 0.5827 | |
| DNA-GPT(Neo-2.7)Interference=DIPPER2023.05 | 0.5595 | |
| DNA-GPTAuxiliary Model=Vicuna-7b2023.05 | 0.5594 | |
| DNA-GPT(Vicuna-7b)Interference=Original2023.05 | 0.5594 | |
| EntropyInterference=DIPPER2023.05 | 0.5552 | |
| DNA-GPT(Vicuna-7b)Interference=DIPPER2023.05 | 0.5523 | |
| DNA-GPT(ChatGPT)Interference=DIPPER2023.05 | 0.5454 | |
| NPRStrategy=Zero-shot2023.05 | 0.5211 | |
| NPRInterference=Original2023.05 | 0.5211 | |
| DetectGPTInterference=Back-translate2023.05 | 0.453 | |
| EntropyStrategy=Zero-shot2023.05 | 0.4364 | |
| EntropyInterference=Original2023.05 | 0.4364 | |
| EntropyInterference=Back-translate2023.05 | 0.3705 | |
| DetectGPTStrategy=Zero-shot2023.05 | 0.3408 | |
| DetectGPTInterference=Original2023.05 | 0.3408 | |
| NPRInterference=DIPPER2023.05 | 0.3006 | |
| DetectGPTInterference=DIPPER2023.05 | 0.1823 |