BidirLM/colpali_train_retrieval

该数据集是用于训练BidirLM-Omni模型的语料库中特定模态的子样本。如需访问完整的、平衡的180万样本全模态数据集（整合文本、图像、音频），请访问全局集成中心：BidirLM/BidirLM-Omni-Contrastive。This dataset is a modality-specific subsample of the corpus employed to train the BidirLM-Omni model. To obtain access to the complete and balanced full-modality dataset containing 1.8 million samples (integrating text, images, and audio), please visit the Global Integration Hub: BidirLM/BidirLM-Omni-Contrastive.