InfographicVQA: Results as Published

InfographicVQA asks questions about infographic images, which mix writing, graphics, layout, and charts. The paper reports 30,035 questions on 5,485 images, including items that need counting, sorting, or arithmetic, and answers that are a text span or a number. A prediction is scored with Average Normalized Levenshtein Similarity, which still gives partial credit for a small string edit, and also with exact-match accuracy. The similarity gives no credit once the normalized edit distance reaches 0.5.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.