Document Image Question Answering on InfographicVQA (test) benchmark leaderboard · SOTA2 Research

Document Image Question Answering on InfographicVQA (test)

0.612ANLSTILT

Evaluation Results

Method	Links
TILT2023.06		0.612	—	67.65	64.19	43.91	38.32	59.17	79.16	45.45	56.54	44.8	48.01	49.58	26.52
Claude + LATIN-Prompt2023.06		0.5451	0.4653	59.92	58.61	39.85	35.44	54.21	67.25	48.97	50.27	49.82	45.98	43.11	27.08
ChatGPT-3.5 + LATIN-Prompt2023.06		0.4898	0.1563	54.57	56.39	34.58	27.98	49.17	60.16	44.91	45.85	36.14	43.12	31.57	26.6
pix2struct2023.06		0.4001	—	43.08	48.39	20.59	31.73	38.33	52.56	25.72	37.26	32.83	27.62	41.98	20.17
ChatGPT-3.5 + Plain Prompt2023.06		0.3335	—	37.49	45.05	9.5	18.22	34.81	38.93	36.7	31.14	18.43	23.49	14.66	23.2
BROS2023.06		0.3219	—	39.97	23.17	10.64	10.68	26.53	44.88	18.78	30.95	32.31	20.2	14.8	6.95
LayoutLMv22023.06		0.2829	—	34.3	27.63	6.41	11.14	24.49	38.55	14.4	26.01	31.1	18.97	11.3	11.58
LayoutLM2023.06		0.272	—	32.78	23.86	4.5	13.71	24	36.26	17.05	25.51	22.05	18.36	15.59	11.4
BERT2023.06		0.2078	—	26.25	23.33	7.39	2.59	18.52	29.95	8.96	19.42	17.09	18.05	1.6	4.36
Claude + Plain Prompt2023.06		0.0798	—	9.51	9.13	2.03	2.8	8.49	10.99	8.58	6.95	4.96	5.89	2.71	3.68