Binary Classification on Apple-defined Content Rating Descriptors

87.62R+ Metric ValueGemini-2.5-Flash

Evaluation Results

Method
Gemini-2.5-Flash2026.05	87.62	97.89
QwenSafe2026.05	74.36	98.62
LLaVA-v1.6-mistral-7b-hf2026.05	74.29	97.01
LLaVA-v1.6-mistral-7b-hf2026.05	71.79	98.85
Gemini-2.5-Flash2026.05	71.79	98.38
QwenSafe2026.05	60.95	93.84
QwenSafe2026.05	60.27	96.64
QwenSafe2026.05	57.66	94.33
Gemini-2.5-Flash2026.05	55.97	88.69
Gemini-2.5-Flash2026.05	55.86	92.41
QwenSafe2026.05	55.48	79.78
QwenSafe2026.05	54.22	91.15
Gemini-2.5-Flash2026.05	52.3	75.89
QwenSafe2026.05	52.2	91.1
Gemini-2.5-Flash2026.05	52.05	94.78
QwenSafe2026.05	51.45	92.29
Qwen3-VL-8B2026.05	51.28	98.02
QwenSafe2026.05	51.05	92.63
Gemini-2.5-Flash2026.05	50.01	90.34
Gemini-2.5-Flash2026.05	48.7	90.95
QwenSafe2026.05	46.28	88.55
LLaVA-v1.6-mistral-7b-hf2026.05	45.22	93.22
LLaVA-v1.6-mistral-7b-hf2026.05	45.05	93.45
Gemini-2.5-Flash2026.05	43.37	85.22
Gemini-2.5-Flash2026.05	42.75	87.83
Gemini-2.5-Flash2026.05	41.49	82.54
LLaVA-v1.6-mistral-7b-hf2026.05	40.54	97.72
QwenSafe2026.05	40	92.38
LLaVA-v1.6-mistral-7b-hf2026.05	37.52	91.15
Qwen3-VL-8B2026.05	36.94	92.55
QwenSafe2026.05	35.14	97.44
LLaVA-v1.6-mistral-7b-hf2026.05	34.25	94.98
LLaVA-v1.6-mistral-7b-hf2026.05	33.96	88.67
LLaVA-v1.6-mistral-7b-hf2026.05	33.33	90.16
Gemini-2.5-Flash2026.05	32.43	96.41
Qwen3-VL-8B2026.05	28.3	87.82
Qwen3-VL-8B2026.05	27.54	89.4
Qwen3-VL-8B2026.05	26.09	91.06
Qwen3-VL-8B2026.05	26.03	94.39
QwenSafe2026.05	24.56	95.46
Qwen3-VL-8B2026.05	24.38	76.54
Qwen3-VL-8B2026.05	24.1	89.99
LLaVA-v1.6-mistral-7b-hf2026.05	22.87	84.54
LLaVA-v1.6-mistral-7b-hf2026.05	16.87	85.52
LLaVA-v1.6-mistral-7b-hf2026.05	16.25	74.65
Qwen3-VL-8B2026.05	16.22	96.82
Qwen3-VL-8B2026.05	16.19	90.87
LLaVA-v1.6-mistral-7b-hf2026.05	15.79	95.06
Gemini-2.5-Flash2026.05	15.79	93.13
Qwen3-VL-8B2026.05	13.86	85.07
Qwen3-VL-8B2026.05	12.28	94.87
Qwen3-VL-8B2026.05	10.11	82.43