Image Captioning on DID-Bench GT-GPT4-V 1.0 (test) benchmark leaderboard · SOTA2 Research

Image Captioning on DID-Bench GT-GPT4-V 1.0 (test)

36.83BLEU-1Patch Matters

Evaluation Results

Method	Links
Patch Matters2025.04		36.83	19	9.58	5.11	0.275	20.01	19.49	19.67	43.46
IT2025.04		33.73	17.39	8.92	4.81	0.187	17.5	19.09	19.46	43.04
Patch Matters2025.04		30.96	16.68	8.73	4.81	0.276	16.73	19.35	20.24	42.47
Synthesized2025.04		27.7	14.29	7.18	3.79	0.112	16.04	19.07	19.45	43.01
Patch Matters2025.04		22.82	11.89	6.1	3.24	0.211	14.67	17.82	19.57	41.87
IT2025.04		21.35	11.65	6.24	3.54	0.003	11.97	17.04	18.58	39.91
Synthesized2025.04		20.52	11.02	5.71	3.12	0.013	13.92	18.94	20.1	41.78
Global2025.04		19.52	9.58	4.76	2.54	0.209	12.81	14.81	15.02	38.89
IT2025.04		16.46	9.03	4.92	2.83	0.002	12.89	17.28	19.26	40.59
Synthesized2025.04		15.77	8.42	4.45	2.45	0.012	12.96	17.67	19.32	41.86
Global2025.04		7.7	4.12	2.15	1.21	0.042	9.9	15.52	17.34	39.05
Global2025.04		7.01	3.72	1.98	1.1	0.004	9.95	14.58	17.8	39.44
PoCa2025.04		4.43	2.3	1.15	0.62	0	8.56	13.19	15.09	38.35
PoCa2025.04		1.57	0.83	0.43	0.24	0	7.07	12.32	14.71	37.12
PoCa2025.04		0.58	0.32	0.17	0.09	0	6	11	14.05	36.16