AI and LLMs · Senior · Question 8 of 75

What is benchmark contamination?

Show the answer

Correct answer: Test questions leaked into the model's training data

Explanation

A contaminated model may memorize answers, so public benchmark scores overstate ability. Private, fresh evals are more trustworthy.

Further reading

Problem with a question

AI and LLMs · Senior · question 8

Test yourself

Take the AI and LLMs · Senior quiz: questions and options are shuffled, and you get a review of your mistakes at the end.

Take this quiz