Safety Evaluation on MEMEVOBENCH QA Style 1.0 (test) benchmark leaderboard · SOTA2 Research

Safety Evaluation on MEMEVOBENCH QA Style 1.0 (test)

16ASR (Rd.1)GPT-5

Evaluation Results

Method	Links
GPT-52026.04		16	22	22
GPT-52026.04		16	44	46
Gemini-2.5-Pro2026.04		19	13	14
Gemini-2.5-Pro2026.04		19	23	30
Qwen3-235B-A22B2026.04		20.4	18.5	16.7
Qwen3-235B-A22B2026.04		21.6	28.7	26.9
Claude-3.7-Sonnet2026.04		27	26	26
Claude-3.7-Sonnet2026.04		27	32	35
GPT-52026.04		29	35	46
GPT-52026.04		29	60	65
Llama-3.3-70B2026.04		41	40	34
Llama-3.3-70B2026.04		41	42	49
Gemini-2.5-Pro2026.04		42	49	66
Qwen3-Next-80B2026.04		43.5	25.9	25.9
Qwen3-Next-80B2026.04		45.3	32.4	41.7
Gemini-2.5-Pro2026.04		46	33	32
GPT-52026.04		49	67	59
GPT-52026.04		49	73	78
Qwen3-235B-A22B2026.04		49.1	42.4	45.4
Qwen3-235B-A22B2026.04		49.1	50	53.7
DeepSeek-V3.22026.04		52	55	61
DeepSeek-V3.22026.04		53	45	43
Llama-3.3-70B2026.04		54	62	65
Llama-3.3-70B2026.04		55	55	52
Claude-3.7-Sonnet2026.04		55	51	43
Claude-3.7-Sonnet2026.04		55	57	64
DeepSeek-V3.22026.04		56	50	50
DeepSeek-V3.22026.04		56	58	62
GPT-4o2026.04		58.9	58.3	61.1
GPT-4o2026.04		61.1	70.4	76.9
Qwen3-Next-80B2026.04		64.8	63.8	61.1
Qwen3-32B2026.04		65.7	74.1	76
Qwen3-32B2026.04		65.7	83.3	82.4
Qwen3-Next-80B2026.04		65.7	71.3	72.3
Gemini-2.5-Pro2026.04		66	72	80
Gemini-2.5-Pro2026.04		67	55	55
Claude-3.7-Sonnet2026.04		69	69	63
Claude-3.7-Sonnet2026.04		69	75	77
GPT-4o2026.04		72.2	79.6	79
GPT-4o2026.04		72.2	84.3	84.3
Qwen3-32B2026.04		72.3	84.3	82.4
Qwen3-32B2026.04		72.3	93.5	94.3
Qwen3-235B-A22B2026.04		76.9	77.8	76.9
Qwen3-235B-A22B2026.04		76.9	86.1	83.1
Llama-3.3-70B2026.04		82	84	81
Llama-3.3-70B2026.04		83	87	90
DeepSeek-V3.22026.04		84	83	81
Qwen3-Next-80B2026.04		84.2	80.5	79.6
Qwen3-Next-80B2026.04		84.2	88.9	84.4
DeepSeek-V3.22026.04		85	86	89
Qwen3-32B2026.04		85.2	91.7	89.8
Qwen3-32B2026.04		86.1	96.3	94.4
GPT-4o2026.04		87	93.5	94
GPT-4o2026.04		88.9	89.8	91.7