MLVU: Results as Published

MLVU, Multi-task Long Video Understanding, asks a model questions about long videos whose lengths run from 3 minutes to more than 2 hours, including movies, surveillance, egocentric video, cartoons, and games. The paper defines 9 tasks, among them topic reasoning, anomaly recognition, summarization, and needle question answering, and it counts 1,730 videos and 3,102 questions. Multiple-choice tasks are scored by accuracy. Open-ended generation is scored by a judge model, and the paper averages the two groups separately as M-Avg and G-Avg.

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.