ViDoRe-v-2

ViDoRe V-2 is a benchmark dataset used to evaluate text-image retrieval models, assessing their performance in retrieving relevant images based on textual queries.

Papers using ViDoRe-v-2 (6)

Llama Nemoretriever Colembed: Top-performing Text-image Retrieval Model2025 · 1 citesReinpool: Reinforcement Learning Pooling Multi-vector Embeddings For Retrieval System2026Evo-Retriever: LLM-Guided Curriculum Evolution with Viewpoint-Pathway Collaboration for Multimodal Document Retrieval2026Structural Anchor Pruning: Training-Free Multi-Vector Compression for Visual Document Retrieval2026MoCa: Modality-aware Continual Pre-training Makes Better Bidirectional Multimodal Embeddings2025Colmate: Contrastive Late Interaction And Masked Text For Multimodal Document Retrieval2025