文献分享: PLAID——为ColBERT架构设计的后期交互驱动器
👉前情提要：
- 神经网络自然语言模型概述
- Transformer \text{Transformer} Transformer与注意力机制概述
📚相关论文：
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding \text{BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding} BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
- 提出了基于双向深度 Transformer \text{Transformer} Transformer的 BERT \text{BERT} BERT交叉编码器
- BERT \text{BERT} BERT的总结
- ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT \text{ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT} ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
- 提出了基于 BERT \text{BERT} BERT编码的后期 Token \text{Token} Token级交互模式
- ColBERTv1 \text{ColBERTv1} ColBERTv1的总结
- ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction \text{ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction} ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction
- 保留了 ColBERT \text{ColBERT} ColBERT的后期交互架构，但从训练策略 / / /嵌入压缩 / / /数据集上优化
- ColBERTv2 \text{ColBERTv2} ColBERTv2的总结
- PLAID: An Efficient Engine for Late Interaction Retrieval \text{PLAID: An Efficient Engine for Late Interaction Retrieval} PLAID: An Efficient Engine for Late Interaction Retrieval
- 在 ColBERTv2 \text{ColBERTv2} ColBERTv2的基础上，进一步改进了检索策略
- PLAID \text{PLAID} PLAID的总结
- EMVB: Efficient Multi-Vector Dense Retrieval Using Bit Vectors \text{EMVB: Efficient Multi-Vector Dense Retrieval Using Bit Vectors} EMVB: Efficient Multi-Vector Dense Retrieval Using Bit Vectors