Benign Task Utility on DriveVLM-Large benchmark leaderboard · SOTA2 Research

Benign Task Utility on DriveVLM-Large

67.22BLEU-1GLA

Evaluation Results

Method	Links
GLA2026.04		67.22	60.87	55.4	50.35	36.92	69.6	3.05
GLA2026.04		64.73	58.33	52.8	47.72	35.23	69.8	3.05
BadNets2026.04		60.13	57.31	50.12	46.51	34.01	68.5	3.01
Blended2026.04		60.11	55.52	46.07	47.33	33.28	68.3	2.98
ISSBA2026.04		60.01	58.92	54.31	50.33	35.89	68.31	3.02
ISSBA2026.04		59.34	59.32	51.34	47.62	34.13	69.1	3.01
ISSBA2026.04		58.92	52.45	46.15	40.3	31.95	65.52	2.9
GLA2026.04		58.73	52.48	47.17	42.3	31.96	67.9	2.91
BadNets2026.04		57.08	51.85	44.31	41.11	30.89	67.7	2.75
w/o attack2026.04		57.07	50.53	45.05	40.11	34.34	70.72	3.1
Blended2026.04		57.01	52.01	45.52	39.25	30.95	66.51	2.89
Blended2026.04		56.33	54.71	52.62	41.44	34.27	67.55	2.91
BadNets2026.04		56.25	55.42	53.61	45.43	33.89	68.1	3