List your company

ResearchFormerly WizWand

AI Detector Attack Effectiveness Evaluation on RAID

96MAGE ASRMAGE BERTScore 0

Evaluation results

Method	Links
MAGE BERTScore 02026.05		96	0.821	27	98	30	89	14	99	16	99	6
MAGE BERTScore Linear 0.12026.05		96	0.807	26	100	31	63	29	94	13	96	22
MAGE BERTScore Linear 0.32026.05		86	0.862	14	98	30	74	26	91	12	90	29
DEPO2026.05		86	0.878	11	99	33	65	32	85	15	78	37
RADAR BERTScore Linear2026.05		34	0.783	17	99	21	97	27	99	23	99	14
RADAR BERTScore2026.05		33	0.775	19	99	21	98	29	99	28	100	16
DEPO2026.05		19	0.85	27	98	13	95	9	87	4	84	14
MAGE BERTScore Linear 0.52026.05		12	0.881	26	89	10	20	44	48	39	50	44
AuthorMist2026.05		9	0.825	32	94	7	84	24	91	1	89	9
AuthorMist2026.05		7	0.831	31	98	27	78	20	90	7	96	19
StealthRL2026.05		3	0.733	34	100	3	26	46	99	31	98	20
StealthRL2026.05		3	0.73	33	94	21	22	43	96	19	97	5
Adv. Paraphrase2026.05		2	0.876	37	98	3	77	35	57	27	41	31
Direct Paraphrasing2026.05		1	0.877	38	96	7	72	37	50	29	34	32
Original2026.05		0	—	47	54	35	44	42	23	31	20	32
SilverSpeak2026.05		0	0.374	33	21	48	14	49	25	48	100	23
RADAR BERTScore2026.05		0	0.991	47	56	35	46	41	23	32	19	32
Original2026.05		0	—	46	30	27	10	46	6	48	3	49
Direct Paraphrasing2026.05		0	0.883	38	84	7	35	39	36	42	27	48
SilverSpeak2026.05		0	0.373	33	0	49	0	49	5	49	100	24
Adv. Paraphrase2026.05		0	0.879	48	97	25	99	26	58	40	21	48
MAGE BERTScore 12026.05		0	0.986	45	33	26	12	45	5	48	3	49