What is log loss in classification?

Updated May 15, 2026

Short answer

Log loss measures classification model performance based on probability predictions.

Deep explanation

It penalizes confident wrong predictions heavily using cross-entropy formulation.

Real-world example

Used in Kaggle competitions for classification ranking.

Common mistakes

  • Using hard labels instead of probabilities.

Follow-up questions

  • Why is log loss better than accuracy?

More Classification interview questions

View all →