Boundary Detection on RoFT original (test) benchmark leaderboard · SOTA2 Research

Boundary Detection on RoFT original (test)

49.64AccuracyROBERTa + SEP

Evaluation Results

Method	Links
ROBERTa + SEP2023.11		49.64	79.71	2.63
ROBERTa2023.11		46.47	74.86	3
SROBERTa2023.11		42	—	—
Phi-2 Perpl. + GB classifier2023.11		35.4	58.3	8.8
Phi-1.5 Perpl. + GB classifier2023.11		29.7	55	9.5
Phi-1.5 Perpl. + LR classifier2023.11		27	49.5	11.9
Phi-2 Perpl. + LR classifier2023.11		25.8	49.5	11.78
GPT-2 Perpl. + GB classifier2023.11		24.25	47.23	11.68
GPT-2 Perpl. + LR classifier2023.11		23.75	42.15	15.8
PHD + TS ML2023.11		23.5	46.32	14.14
Human baseline2023.11		22.62	40.31	13.88
LLaMA-2-7B Perpl. + GB classifier2023.11		22.3	42	14.79
Phi-1 Perpl. + GB classifier2023.11		20.9	42.7	12.92
DetectGPT + GB classifier2023.11		19.79	37.4	8.35
DetectGPT + LR classifier2023.11		19.45	33.82	9.03
LLaMA-2-7B Perpl. + LR classifier2023.11		19.4	34.7	19.75
Phi-1 Perpl. + LR classifier2023.11		19.2	38.3	16.26
Phi-1.5 Perpl. + GB regressor2023.11		17.1	44.6	6.11
Phi-2 Perpl. + GB regressor2023.11		15.6	43.7	5.83
Majority class2023.11		15.26	25.43	27.58
Length + GB2023.11		14.64	33.43	16.55
GPT-2 Perpl. + GB regressor2023.11		12.58	36.67	6.89
TLE + TS Binary2023.11		12.58	30.41	22.23
LLaMA-2-7B Perpl. + GB regressor2023.11		11.4	32.9	7.98
Phi-1 Perpl. + GB regressor2023.11		10.7	33	7.51
gpt-42023.11		6.4	22	27
gpt-3.5-turbo2023.11		2.5	17.3	30.3