M-2KR

The M-2KR dataset/benchmark contains multimodal document collections with both images and text, and it is used to evaluate the performance of retrieval models on complex multimodal queries.

Papers using M-2KR (5)

Recurrence-Enhanced Vision-and-Language Transformers for Robust Multimodal Document Retrieval2025 · 4 citesOverview of the EReL@MIR 2025 Multimodal Document Retrieval Challenge (Track 1)2026WikiSeeker: Rethinking the Role of Vision-Language Models in Knowledge-Based Visual Question Answering2026A Multi-Granularity Retrieval Framework for Visually-Rich Documents2025PreFLMR: Scaling Up Fine-Grained Late-Interaction Multi-modal Retrievers2024