Quadruple-consistency evaluation on CCTVBENCH benchmark leaderboard · SOTA2 Research

Quadruple-consistency evaluation on CCTVBENCH

29.07Video Consistency (Contr@q+)GPT-5

Evaluation Results

Method	Links
GPT-52026.04		29.07	87.13	58.1	56.64	55.47	56.05	70.43	48.43	25.85
Gemini-2.52026.04		24.48	83.79	54.13	55.56	48.91	52.24	69.66	46.43	20.87
VideoLLaMA32026.04		21.65	31.16	26.41	27.44	28.28	27.86	55.38	29.89	6.48
Gemini-2.52026.04		21.43	86.95	54.19	59.42	48.86	54.14	70.37	47.71	18.28
InternVL3.52026.04		20.88	64.92	42.9	38.58	48.49	43.53	64.85	39.89	11.38
VideoLLaMA32026.04		20.04	36.8	28.42	36.41	21.81	29.11	58.47	32.91	6.58
Qwen3-VL2026.04		17.02	63.51	40.26	34.21	47.94	41.07	62.96	37.77	7.62
VideoChatGPT2026.04		16.21	21.45	18.83	18.59	19.81	19.2	49.85	24.87	4.13
GPT-52026.04		13.92	91.19	52.55	26.24	68.39	47.31	57.46	34.43	12.6
PLLaVA2026.04		13.87	66.67	40.27	29.06	52.39	40.73	59.81	34.7	6.19
InternVL3.52026.04		13.05	63.79	38.42	24.23	53.54	38.88	58.46	33.17	7.05
Qwen3-VL2026.04		13.01	91.05	52.03	27.85	76.38	52.11	60.26	38.62	10.65
Phi-4-Multimodal2026.04		11.23	40.07	25.65	27.26	25.38	26.32	57.07	31.61	2.92
PLLaVA2026.04		8.89	80.75	44.82	22.35	68.95	45.65	57.38	33.35	3.9