News Classification with LLMs: Building Reliable Evaluation Datasets – Trine Engelund, JP/Politiken Media Group

In this session, Trine Engelund from JP/Politiken Media Group, focuses on how to build reliable evaluation datasets for LLMs.
Data Innovation Summit 2024 Data Innovation Summit 2024
Data Innovation Summit 2024

Session Outline

In this session at the Data Innovation Summit 2024, we have Trine Engelund, Machine Learning Specialist at JP/Politiken Media Group! In her talk, Trine focuses on how to build reliable evaluation datasets for LLMs. The basis of the session is a case study on classifying the topicality of news articles from Ekstra Bladet, Denmark’s largest online news media.

Key Takeaways

  • The importance of evaluating LLMs before deploying, creating an annotation guide, and investing in the annotation process for reliable evaluation.
Add a comment

Leave a Reply