This dataset is associated with the paper Equip Pre-ranking with Target Attention by Residual Quantization.
TARQ is a novel pre-ranking framework designed for industrial recommendation systems. It bridges the gap between efficiency and effectiveness by bringing the modeling power of Target Attention (TA) into the latency-critical pre-ranking stage through Residual Quantization.
Paper: Equip Pre-ranking with Target Attention by Residual… See the full description on the dataset page:
https://huggingface.co/datasets/zyody/TARQ_dataset.