Harmfulness Analysis on AdamMeme 1.0 (test) benchmark leaderboard · SOTA2 Research

Harmfulness Analysis on AdamMeme 1.0 (test)

3.84Nationality Harm ScoreQwen-VL-Chat

Evaluation Results

Method	Links
Qwen-VL-Chat2025.07		3.84	65.73	4.06	51.38	4.44	47.58	4.55	45.91	3.5	67.92	4.06	54.86	4.19	52.65	4.33	49.16	4.13	54.14
Qwen-VL-Max2025.07		4.96	38.77	5.19	29.74	5.07	32.88	4.79	40.64	4.46	49.77	5.16	33.33	4.74	45.57	4.93	38.67	4.92	38.63
Doubao-Pro2025.07		5.1	39.11	4.17	54.92	5.41	33.08	5.16	38.65	4.07	62.01	4.09	58.75	4.8	43.88	4.48	52.23	4.67	47.58
Doubao-Lite2025.07		5.25	40.48	5.61	24.46	5.68	25.28	5.64	26.52	5.34	28.5	6.12	16.6	5.61	30.98	5.29	30.83	5.57	28.02
LLaVA-v1.62025.07		5.28	34.16	5.6	26.25	5.01	35.29	5	37.88	4.37	51.53	4.86	39.3	4.81	47.7	5.46	24.57	5.05	37.06
LLaVA-v1.62025.07		5.9	21.29	6.13	18.8	6.16	16.06	6.05	18.6	5.89	19.18	6.19	17.37	6.01	21.63	6.05	16.32	6.05	18.66
Qwen2.5-VL2025.07		5.99	25.22	6.53	16.6	6.32	18.34	6.45	20.35	5.69	26.57	6.36	18.34	5.79	29.31	6.52	10.05	6.21	20.59
QwQ2025.07		6.19	14.89	6.26	18.6	6.24	18.43	6.16	17.14	5.41	30.52	6.47	10.34	6.03	20	6.28	11.16	6.14	17.53
Step-1v2025.07		6.93	10.92	6.87	9.79	7	4.8	6.89	5.81	6.63	10.23	6.47	17.92	6.86	12.24	6.83	5.83	6.81	9.7
Step-1o-Vision2025.07		7.4	5.33	7.68	3.4	7.68	4.08	7.38	5.22	7.36	2.87	7.29	7	7.46	6.33	7.28	5.22	7.44	4.97
GPT-4o2025.07		7.53	0.43	7.43	2.14	7.52	1.24	7.3	3.53	7.15	3.64	7.44	2.54	7.39	3.36	7.26	0.44	7.38	2.18