Multi-episode Video Understanding on SeriesBench benchmark leaderboard · SOTA2 Research

Multi-episode Video Understanding on SeriesBench

98.2VS ScoreHuman

Evaluation Results

Method	Links
Human2025.04		98.2	94.4	94.6	97.2	92.6	95.8	—	—	—
GPT-4o2025.04		78.6	76.1	73.8	82.1	61.7	76.2	12.61	28.29	72.75
InternVL22025.04		76.5	71.4	67.3	81.1	66.7	73.3	8.26	20.71	67.85
MiniCPM-V 2.62025.04		76.3	68.2	66.5	82.5	60	72	14.16	33.64	72.98
Qwen2-VL2025.04		75.2	73.5	69.7	80.9	67.4	73.9	8.93	23.41	68.08
GPT-4o2025.04		55.8	62.8	60.6	79.9	59.6	62.8	9.61	25.1	68.94
Qwen2-VL2025.04		55.7	57.5	58.6	75.3	59.6	60.3	11.41	27.97	70.71
LLaVA-Video2025.04		54.9	56.8	57.4	71.2	51.9	58.3	7.27	19.06	67.95
MiniCPM-V 2.62025.04		53.3	57.3	57.7	76.6	55.1	59.1	8.57	30.06	68.66
Aria2025.04		52.4	56.6	51.8	75.6	50	56.8	10	26.08	70.35
InternVL22025.04		52	56.8	57.7	78.6	57.7	59.2	8.43	21.22	68.4
LLaVA-OneVision2025.04		51.5	54	56.2	70.6	59.6	56.9	7.3	20.2	67.82
VideoLLaMA2.1-AV2025.04		47.4	52.2	53.3	66.6	49.4	53.1	7.61	23.34	67.53
VITA2025.04		46.9	46	53.3	68.9	48.1	51.8	9.32	27.08	68.93
GPT-4o-mini2025.04		46.7	42.7	47.9	70.9	44.9	49.8	9.99	29.3	68.76
VideoLLaMA2.1-AV2025.04		45.3	52.2	51.3	66.2	46.2	51.7	7.24	22.71	67.09
Frequent Guess2025.04		43.6	46	43.1	41.1	50.6	44.4	—	—	—
Random Choice2025.04		39.3	38.2	35.5	36.5	38.8	37.7	—	—	—