Video-based Dialogue Evaluation on Video-ChatGPT benchmark leaderboard · SOTA2 Research

Video-based Dialogue Evaluation on Video-ChatGPT

3.06CIFULL CACHE

Evaluation Results

Method	Links
FULL CACHE2026.04		3.06	3.15	3.52	2.24	2.69	2.93
HYBRIDKV2026.04		2.99	3.05	3.47	2.15	2.57	2.85
FULL CACHE2026.04		2.99	2.86	3.36	2	2.99	2.84
HYBRIDKV2026.04		2.96	2.88	3.35	1.89	2.89	2.79
MADAKV2026.04		2.94	3.03	3.41	2.02	2.56	2.79
LOOK-M2026.04		2.92	2.97	3.38	2.05	2.49	2.76
SNAPKV2026.04		2.9	2.96	3.4	1.95	2.53	2.75
SPARSEMM2026.04		2.9	2.95	3.37	1.99	2.52	2.75
LOOK-M2026.04		2.9	2.75	3.32	1.87	2.87	2.74
MADAKV2026.04		2.9	2.79	3.35	1.9	2.89	2.77
SNAPKV2026.04		2.88	2.78	3.31	1.83	2.8	2.72
TemporalVLM2024.12		2.88	2.91	3.45	2.5	3.16	—
SPARSEMM2026.04		2.87	2.73	3.27	1.91	2.84	2.72
LongVLM2024.12		2.76	2.86	3.34	2.39	3.11	—
LOOK-M2026.04		2.74	2.68	3.1	1.85	2.23	2.52
HYBRIDKV2026.04		2.7	2.72	3.11	1.84	2.43	2.56
SPARSEMM2026.04		2.69	2.66	3.11	1.82	2.25	2.51
BT-Adapter2024.12		2.68	2.69	3.27	2.34	2.46	—
SNAPKV2026.04		2.64	2.67	3.09	1.77	2.23	2.48
MADAKV2026.04		2.6	2.64	3.03	1.81	2.39	2.49
FULL CACHE2026.04		2.59	2.67	3.06	1.93	2.49	2.55
Video-ChatGPT2024.12		2.5	2.57	2.69	2.16	2.2	—
Valley2024.12		2.43	2.13	2.86	2.04	2.45	—
Video-LLaMA2024.12		1.96	2.18	2.16	1.82	1.79	—