MM-IMDb

The 'MM-IMDb' dataset is a benchmark used to evaluate vision-language models, specifically focusing on their performance and bias in multimodal tasks involving movie-related content.

Papers using MM-IMDb (3)

DAIN: Dynamic Agent-Based Interaction Network for Efficient and Collaborative Multimodal Reasoning2026EPIC: Efficient Prompt Interaction for Text-Image Classification2025See-Saw Modality Balance: See Gradient, and Sew Impaired Vision-Language Balance to Mitigate Dominant Modality Bias2025