MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly
This provides a foundational benchmark for researchers and developers working on long-context vision-language models, enabling systematic evaluation and diagnosis of model capabilities.
The authors tackled the problem of evaluating long-context vision-language models (LCVLMs) by introducing MMLongBench, a comprehensive benchmark with 13,331 examples across five task categories and standardized input lengths up to 128K tokens. Their benchmarking of 46 models revealed that single-task performance is a weak proxy for overall capability, both closed-source and open-source models struggle with these tasks, and models with stronger reasoning ability tend to perform better.
The rapid extension of context windows in large vision-language models has given rise to long-context vision-language models (LCVLMs), which are capable of handling hundreds of images with interleaved text tokens in a single forward pass. In this work, we introduce MMLongBench, the first benchmark covering a diverse set of long-context vision-language tasks, to evaluate LCVLMs effectively and thoroughly. MMLongBench is composed of 13,331 examples spanning five different categories of downstream tasks, such as Visual RAG and Many-Shot ICL. It also provides broad coverage of image types, including various natural and synthetic images. To assess the robustness of the models to different input lengths, all examples are delivered at five standardized input lengths (8K-128K tokens) via a cross-modal tokenization scheme that combines vision patches and text tokens. Through a thorough benchmarking of 46 closed-source and open-source LCVLMs, we provide a comprehensive analysis of the current models' vision-language long-context ability. Our results show that: i) performance on a single task is a weak proxy for overall long-context capability; ii) both closed-source and open-source models face challenges in long-context vision-language tasks, indicating substantial room for future improvement; iii) models with stronger reasoning ability tend to exhibit better long-context performance. By offering wide task coverage, various image types, and rigorous length control, MMLongBench provides the missing foundation for diagnosing and advancing the next generation of LCVLMs.