3D Object Captioning on Cap3D subset of 3186 3D objects (test) benchmark leaderboard · SOTA2 Research

3D Object Captioning on Cap3D subset of 3186 3D objects (test)

0.312CLIP Image-Text ScoreKimi-VL

Evaluation Results

Method	Links
Kimi-VL2026.04		0.312	0.597	0.213	34.93	27.9	13.87	0.116
Bagel2026.04		0.299	0.664	0.15	35.06	28.28	14.63	0.145
Qwen2.5-VL-3B2026.04		0.299	0.661	0.199	34.78	27.81	13.25	0.121
UniMesh2026.04		0.297	0.686	0.113	35.97	28.09	13.72	0.155
Qwen2.5-VL-7B2026.04		0.296	0.715	0.185	32.33	25.77	12.21	0.133
InternVL3.5-4B2026.04		0.295	0.658	0.149	32.74	25.71	12.84	0.16
DiffuRank2026.04		0.291	0.68	0.137	38.61	30.82	16.07	0.133
Cap3D2026.04		0.287	0.694	0.123	41.27	33.52	17.51	0.158
Phi-4-multimodel2026.04		0.285	0.708	0.172	24.48	19.46	9.45	0.172
LLaVA-3D-7B2026.04		0.278	0.698	0.209	20.34	15.57	7.19	0.146