Video Retrieval on RVMS-Bench 1.0 (test) benchmark leaderboard · SOTA2 Research

Video Retrieval on RVMS-Bench 1.0 (test)

64GHuman

Evaluation Results

Method	Links
Human2026.02		64	—	28	24	88	24	56	64	60	80	54.2
Human2026.02		64	—	—	—	88	—	—	72	60	80	72.8
Gemini-3-Pro2026.02		45.6	—	—	—	58.8	—	—	63.1	56.9	61.2	57
Gemini-2.5-Pro2026.02		36.2	—	—	—	28.8	—	—	34.4	30	24.4	31
Qwen3-VL-235B2026.02		32.5	—	—	—	42.5	—	—	40	43.8	25	36.6
GPT-5-Mini2026.02		31.2	—	—	—	27.5	—	—	28.8	28.8	27.5	28.8
Qwen2.5-VL-72B2026.02		29.4	—	—	—	23.1	—	—	33.8	26.2	28.8	28.1
Hunyuan-Video2026.02		28.8	—	—	—	22.5	—	—	27.5	23.8	19.4	24.4
GPT-5.22026.02		26.2	—	—	—	36.9	—	—	38.1	31.2	36.2	33.8
Claude-4.5-Sonnet2026.02		23.8	—	—	—	24.4	—	—	26.2	21.9	23.8	24
Qwen3-VL-235B2026.02		21.9	—	15.1	12.4	13.1	15	16.1	11.9	23.1	30	17.6
Gemini-2.5-Pro2026.02		18.8	—	13.1	10.6	15	12.5	13.8	18.1	16.2	17.5	15.1
GPT-4o2026.02		18.8	—	—	—	19.4	—	—	21.9	17.5	18.8	19.2
Qwen2.5-VL-72B2026.02		16.9	—	11.9	10	11.9	11.2	12.5	17.5	15	16.2	13.7
GPT-5-Mini2026.02		16.1	—	1.9	1.9	10.6	4.4	4	7	13.8	13.1	8.1
Hunyuan-Video2026.02		15.6	—	10.6	8.8	12.5	10	11.2	15	13.1	14.4	12.4
Gemini-3-Pro2026.02		14.4	—	23.8	16.9	16.9	20	23.8	31.9	15.6	22.5	20.6
GPT-5.22026.02		14.4	—	3.2	1.6	12.9	6.9	5	13.1	16.2	15	9.8
Claude-4.5-Sonnet2026.02		12.5	—	10.6	8.8	12.5	10	11.2	14.4	11.2	13.1	11.6
GPT-4o2026.02		10.6	—	7.5	6.2	10	7.5	8.1	11.9	10	11.2	9.2