Share your thoughts, 1 month free Claude Pro on usSee more

Scene Captioning on nocaps RGBP seen scenes (val)

102.67CIDEr

Qwen3-VL-8B-Instruct RGBP

Updated 17d ago

Evaluation Results

Method	Links
Qwen3-VL-8B-Instruct RGBP 2026.04		102.67	13.19
ProCap 2026.04		42.06	6.66
ProCap 2026.04		39.93	6.59
Qwen3-VL-8B-Instruct RGBP 2026.04		37.48	23.49
ProCap 2026.04		37.14	22.75
ProCap 2026.04		36.39	22.88
ProCap 2026.04		35.21	24.21
ProCap 2026.04		30.58	5.17
ProCap 2026.04		30.45	4.8
ProCap 2026.04		27.63	24.45
Qwen3-VL-8B-Instruct 2026.04		15.57	10.01
Qwen3-VL-4B-Instruct 2026.04		14.93	10.47
Qwen3-VL-2B-Instruct 2026.04		14.85	10.38
FastVLM-7B 2026.04		7.01	6.12
FastVLM-0.5B 2026.04		6.16	5.68
FastVLM-1.5B 2026.04		5.72	5.88
FastVLM-7B 2026.04		2.21	10.79
Qwen3-VL-4B-Instruct 2026.04		2.19	11.94
Qwen3-VL-2B-Instruct 2026.04		2.12	12.17
Qwen3-VL-8B-Instruct 2026.04		2.09	13.15
FastVLM-1.5B 2026.04		1.73	9.16
FastVLM-0.5B 2026.04		1.6	9.47