Video Question Answering on HD-EPIC VQA 1.0 (test) benchmark leaderboard · SOTA2 Research

Video Question Answering on HD-EPIC VQA 1.0 (test)

84Following Activity RecognitionEvidence-guided reasoning framework

Updated 2mo ago

Evaluation Results

Method
Evidence-guided reasoning framework2026.05	84	65.8	84	98	78	86	90	88	90	88	92	76	85	76	54	50	82	86	67.6	75.4	41.7	76.8	46	48.2	64	58.5	49	51	58	61.3	38.9
DeepFrames2026.05	70	—	72	70	50	34	73	68	81	76	32	36	48	30	38	25	26	60	36.6	51	24.9	55.6	29	34.2	49.8	50.5	14.2	44.5	31	51.4	14.5
VideoLLaMA22026.05	64	—	52	18	13	22	21	38	13	19	30	20	27	26	32	24	20	54	25.2	32.2	20.7	30.9	17.7	18.8	31	35.5	11	44	30.5	30	12.4
LLaVA-Video2026.05	62	—	68	44	21	28	24	20	23	22	36	38	41	36	28	28	26	62	41.4	51.2	20.9	58.6	16.3	21.8	30.6	40.5	9.8	20	27	47.5	11.1
Gemini Pro2026.05	54	—	76	88	35	42	74	70	45	49	46	56	62	36	28	26	16	62	35.6	43.2	30.3	49.3	35.3	20.8	32.4	41.5	18	13	31.5	36.5	20.8