Multimodal Long-document Understanding on MMLongBench-Doc 1.0 (test) benchmark leaderboard · SOTA2 Research

Multimodal Long-document Understanding on MMLongBench-Doc 1.0 (test)

46.7Reports ScoreBayesRAG

Evaluation Results

Method	Links
BayesRAG2026.01		46.7	42.6	41.8	53.6	32.7	34.8	41.6	46.4	37	39.4	31.1	38.6	41.2	48.6	38.8	44.1
RAGAnything2026.01		45.3	40.6	39.5	46	32	34.6	37.3	43.5	36.4	39.6	30.5	33.3	39.2	42.7	37.1	41.5
AutothinkRAG2026.03		44.86	—	—	52.2	—	52.26	—	52.34	—	51	—	62.19	—	56.14	—	51.29
Nothink2026.03		43.15	—	—	48.55	—	43.21	—	52.25	—	46	—	53.08	—	50.42	—	47.21
8B2026.03		42.8	—	—	44.12	—	43.21	—	52.34	—	46	—	50.61	—	50.42	—	46.09
VLM only2026.03		41.43	—	—	49.27	—	43.43	—	52.25	—	41	—	54.32	—	48.72	—	45.55
Raganything2026.03		41.09	—	—	41.3	—	44.22	—	45.16	—	40	—	54.32	—	52.13	—	44.36
baseline2026.03		40.41	—	—	47.82	—	42.42	—	50.96	—	42	—	51.85	—	46.15	—	44.86
RAGFlow2026.01		38.9	36.8	31.5	35.2	27.8	29.9	30.6	35.2	33.4	36.6	28.7	30.8	31.2	34.1	32	34.9
GPT-4o-mini2026.01		36.7	29.8	29	38.2	16.8	24	24	32.4	29.5	34.6	41.3	50.6	38.3	41	28.2	35.2
ViDoRAG2026.01		36.1	29.8	33.3	39.5	16.5	21	27.6	32	19.6	24.7	21.8	25.9	35.3	36.7	26.5	31.4
VisRAG2026.01		30	27.9	19.4	20.8	21.5	24	19.8	23.7	19.8	23.7	19.7	24.6	11.1	11.1	21.3	23.8