What is a Cross-Encoder?
A type of encoder that jointly processes query-document pairs to determine relevance.
More about Cross-Encoder:
Cross-Encoders take a query and a document as input and process them together through a single model to calculate relevance. Unlike bi-encoders, which generate embeddings independently, cross-encoders allow for finer-grained relevance scoring by considering the interaction between query and document.
Cross-encoders are often used in tasks requiring high accuracy, such as re-ranking candidates retrieved by dense retrieval or hybrid search.
Frequently Asked Questions
What are the advantages of cross-encoders?
They provide higher relevance accuracy by analyzing query-document interactions directly.
When should cross-encoders be used over bi-encoders?
Cross-encoders are better for re-ranking results, while bi-encoders are more efficient for large-scale retrieval.
From the blog
AI Chatbots for Ecommerce: Reducing Cart Abandonment with 24/7 Support
An AI chatbot for ecommerce can help reduce the demand on the support team, offer 24/7 customer support, and boost conversions. See how here.
Herman Schutte
Founder
How to Train ChatGPT With Your Own Website Data
Training ChatGPT with your own data can provide the model with a better understanding of your unique context, allowing for more accurate and relevant responses.
Herman Schutte
Founder