Precision, recall, and threshold tradeoffs (step 6/7) · metrics, slices, and error analysis

promptdojo_

A model gives each request a fraud score from 0 to 1. Write apply_threshold(scores, threshold) that returns a list of labels: "fraud" if the score is >= threshold, otherwise "legit".

This is the dial from the lesson: the same scores produce more or fewer fraud flags depending on the threshold you pass in.

A model gives each request a fraud score from 0 to 1. Write apply_threshold(scores, threshold) that returns a list of labels: "fraud" if the score is >= threshold, otherwise "legit".

This is the dial from the lesson: the same scores produce more or fewer fraud flags depending on the threshold you pass in.

full-screen editor opens — close anytime to keep reading.

Precision, recall, and threshold tradeoffs — step 6 of 7

Precision, recall, and threshold tradeoffs — step 6 of 7