MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly
[Sept 2025] 🎉 MMLongBench is accepted as a spotlight at NeurIPS 2025!!!
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… See the full description on the dataset page: https://huggingface.co/datasets/ZhaoweiWang/MMLongBench.
