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

Image Captioning on COMPOSITIONCAP (test)

47.5ROUGE-L

Qwen2.5-VL-7B + TDSR

Updated 3mo ago

Evaluation Results

Method	Links
Qwen2.5-VL-7B + TDSR 2025.10		47.5	19.7	27.5	129.4	88.9
LLaVA-1.5-7B + TDSR 2025.10		44.3	16.6	23.5	124.2	82.5
Qwen2.5-VL-7B 2025.10		41.2	14.5	21.9	120.3	81.3
VistaLLM-13B 2025.10		40.9	14.1	21.4	117.5	80.2
FINECAPTION-8B 2025.10		40.6	13.9	20.9	118.6	79.5
Alpha-CLIP-13B 2025.10		35.6	10.9	19.3	93.9	77.7
LLaVA-1.5-7B+Patch Matters 2025.10		34.6	12.5	21.2	118.8	79.1
Ferret-13B 2025.10		33.6	12.8	19.6	114.6	79.1
LLaVA-1.5-7B+IT 2025.10		32.9	10.6	15.7	95.2	78.2
Shikra-13B 2025.10		32.4	11.9	19.5	108.4	78.4
MiniGPT-v2-7B 2025.10		31.9	11.5	18.7	106.2	78.2
VisionLLM-H-7B 2025.10		31.2	10.7	15.4	90.2	76.5
KOSMOS-2 2025.10		30.8	10.1	14.9	88.9	76.7
LLaVA-1.5-7B 2025.10		30.3	8.6	11.4	86.5	73.2