Share your thoughts, 1 month free Claude Pro on usSee more

Streaming Video Understanding on OVO-Bench RealStreamEval protocol

93.9OCR

Human Evaluation

Updated 22d ago

Evaluation Results

Method	Links
Human Evaluation 2026.05		93.9	92.5	94.8	92.7	91	94	93.2	92.5	93	91.4	92.3	95.4	89.6	93.5	92.8	92.8
EvoStreaming-8B-Qwen2 2026.05		82.9	67.9	75.9	53.4	62.4	69	68.6	51.5	55.4	45.2	50.7	20.6	64.6	6.7	30.6	54.6
EvoStreaming-8B-Qwen3 2026.05		79.9	50.5	62.1	37.1	38.6	60.6	54.8	45.1	41.2	33.6	40	19.3	69.3	24.2	37.6	46.8
EvoStreaming-8B-Qwen2.5 2026.05		78.9	58.3	63.4	51.1	56.9	60.9	61.6	45.3	53	35.2	44.5	18.9	38.7	7.9	21.8	47.4
Qwen2.5-VL-8B 2026.05		78.5	47.7	58.6	38.2	56.4	58.2	56.3	41.6	56.8	42.5	46.9	14.9	15	8	12.6	43
Qwen2-VL-8B 2026.05		75.2	58.3	61.2	33.7	50	59.8	56.4	48.2	44.3	57	49.8	16	21.2	0.8	12.6	43.8
InternVL-3.5-8B 2026.05		65.8	51.4	52.6	40.2	58.4	53.8	53.7	46	52.7	21	39.9	16.9	22.9	4	14.6	40.5
EvoStreaming-8B-InternVL3.5 2026.05		59.2	51.3	47	34.7	52	48.1	48.7	43.5	49.3	21.2	38	26.1	62.9	29.1	39.4	43.7
Qwen3-VL-8B 2026.05		54.7	44.5	57.3	43.8	35.2	56	48.6	34.2	50.3	38.2	40.9	17.1	24.6	5.6	15.8	38.5
LLaVA-Video-7B 2026.05		50.3	52.3	49.6	31.5	51.5	41.3	46.1	38.4	51.4	16.1	35.3	17.8	23.2	4.9	15.3	35.7
LLaVA-OneVision-7B 2026.05		47.3	50.5	63.8	46.1	51.5	54.1	52.2	37.5	49.3	67.7	51.5	13	20.2	1.2	11.5	41.9
Dispider-7B 2026.05		47	44	45.7	37.1	58.4	48.4	46.8	46.5	51.4	3.2	33.7	5	0	12.4	5.8	33.3
EvoStreaming-8B-MiniCPM-V 2026.05		46.3	45	41.4	29.2	45.5	42.4	41.6	41.1	46	21.5	36.2	23.1	71.4	23.3	39.3	39.7
MiniCPM-V4.5-8B 2026.05		43	48.6	47.8	36	38.1	41.6	42.5	33.2	55.7	43.8	44.2	16.2	23.1	5.4	14.9	34.4
Flash-VStream-7B 2026.05		24.2	29.4	28.5	33.7	25.7	28.8	28.4	39.1	37.2	5.9	27.4	17.4	13.5	5.1	12	24
TimeChat-Online 2026.05		24.2	30.3	31	25.3	48.5	36.4	32.6	37.7	41.9	19.9	33.2	14.7	17.3	3.3	11.8	27.5
Video-LLaMA2-7B 2026.05		20.8	28.4	17.7	15.7	29.7	24.5	22.8	14.8	30.4	7.5	17.6	13.6	21.6	1.3	12.2	18.9