[KongchangAI]
Conceptanswer balance

Answer Balance

A quality-control technique commonly used when constructing multiple-choice or true-false benchmarks. It requires the various answer labels to be roughly matched in distribution, so that the expected accuracy of random guessing remains around 50 percent, ensuring the benchmark's discriminative power.

Timeline (last 90 days)

Oct 15

Answer balancing requires that the yes/no labels of each question type be roughly matched in distribution, so that the expected accuracy of random guessing stays around 50%.

Unverified50%

All Facts (1)

Source Articles