AI-generated text detection on XSum Generated by Claude3 (test)
100AUROCDPIC
Evaluation Results
| Method | Links | |
|---|---|---|
| DPICAuxiliary Model=ChatGPT2023.05 | 100 | |
| DPIC(ChatGPT)Interference=Original2023.05 | 100 | |
| DPIC(ChatGPT)Interference=DIPPER2023.05 | 99.96 | |
| DPIC(Vicuna-7b)Interference=DIPPER2023.05 | 99.96 | |
| DPICAuxiliary Model=Vicuna-7b2023.05 | 99.92 | |
| DPIC(Vicuna-7b)Interference=Original2023.05 | 99.92 | |
| DPIC(ChatGPT)Interference=Back-translate2023.05 | 99.79 | |
| DPIC(Vicuna-7b)Interference=Back-translate2023.05 | 99.66 | |
| RADARStrategy=Supervised2023.05 | 99.52 | |
| RADARInterference=Original2023.05 | 99.52 | |
| Fast-DetectGPTStrategy=Zero-shot2023.05 | 99.42 | |
| Fast-DetectGPTInterference=Original2023.05 | 99.42 | |
| RADARInterference=DIPPER2023.05 | 99.4 | |
| Fast-DetectGPTInterference=Back-translate2023.05 | 98.6 | |
| LogRankStrategy=Zero-shot2023.05 | 97.83 | |
| LogRankInterference=Original2023.05 | 97.83 | |
| LikelihoodStrategy=Zero-shot2023.05 | 97.6 | |
| LikelihoodInterference=Original2023.05 | 97.6 | |
| Fast-DetectGPTInterference=DIPPER2023.05 | 97.2 | |
| RADARInterference=Back-translate2023.05 | 97.01 | |
| LRRStrategy=Zero-shot2023.05 | 96.09 | |
| LRRInterference=Original2023.05 | 96.09 | |
| LogRankInterference=Back-translate2023.05 | 94.88 | |
| LikelihoodInterference=Back-translate2023.05 | 94.46 | |
| LogRankInterference=DIPPER2023.05 | 92.56 | |
| LRRInterference=Back-translate2023.05 | 92.43 | |
| LRRInterference=DIPPER2023.05 | 92.4 | |
| NPRStrategy=Zero-shot2023.05 | 92.32 | |
| NPRInterference=Original2023.05 | 92.32 | |
| LikelihoodInterference=DIPPER2023.05 | 90.8 | |
| DNA-GPTAuxiliary Model=Neo-2.72023.05 | 90.71 | |
| DNA-GPT(Neo-2.7)Interference=Original2023.05 | 90.71 | |
| NPRInterference=Back-translate2023.05 | 90.42 | |
| RoBERTa-largeBackbone=RoBERTa-large2023.05 | 90.27 | |
| ROBERTa-largeInterference=Original2023.05 | 90.27 | |
| RoBERTa-baseBackbone=RoBERTa-base2023.05 | 89.44 | |
| ROBERTa-baseInterference=Original2023.05 | 89.44 | |
| DNA-GPTAuxiliary Model=ChatGPT2023.05 | 85.6 | |
| DNA-GPT(ChatGPT)Interference=Original2023.05 | 85.6 | |
| DNA-GPT(Neo-2.7)Interference=Back-translate2023.05 | 83.99 | |
| ROBERTa-largeInterference=DIPPER2023.05 | 81.53 | |
| ROBERTa-baseInterference=DIPPER2023.05 | 81.2 | |
| ROBERTa-baseInterference=Back-translate2023.05 | 80.52 | |
| DNA-GPT(ChatGPT)Interference=DIPPER2023.05 | 79.96 | |
| DNA-GPT(Neo-2.7)Interference=DIPPER2023.05 | 78.76 | |
| NPRInterference=DIPPER2023.05 | 78.6 | |
| DNA-GPT(ChatGPT)Interference=Back-translate2023.05 | 78.14 | |
| ROBERTa-largeInterference=Back-translate2023.05 | 75.83 | |
| DNA-GPTAuxiliary Model=Vicuna-7b2023.05 | 72.41 | |
| DNA-GPT(Vicuna-7b)Interference=Original2023.05 | 72.41 | |
| DNA-GPT(Vicuna-7b)Interference=DIPPER2023.05 | 64.03 | |
| DNA-GPT(Vicuna-7b)Interference=Back-translate2023.05 | 63.21 | |
| DetectGPTInterference=Back-translate2023.05 | 52.73 | |
| EntropyInterference=DIPPER2023.05 | 49.78 | |
| DetectGPTStrategy=Zero-shot2023.05 | 43.23 | |
| DetectGPTInterference=Original2023.05 | 43.23 | |
| EntropyStrategy=Zero-shot2023.05 | 41.09 | |
| EntropyInterference=Original2023.05 | 41.09 | |
| EntropyInterference=Back-translate2023.05 | 36.39 | |
| DetectGPTInterference=DIPPER2023.05 | 32.83 |