AI engineering glossary · Evals guardrails AI security
What is verbosity bias?
Verbosity bias (also called length bias): A tendency of automated judges to rate longer, more elaborate answers higher even when the extra length adds no substance or adds errors. It can push systems to become wordy. Countermeasures include rubrics that penalise padding, length-controlled scoring and comparing against human preferences.
Verbosity bias: an example
A judge scores a 400-word answer higher than a correct 60-word one to the same question about an EMI calculation, so the team adds a rubric line: 'concise and complete beats long'.
Related terms
More Evals guardrails AI security terms
- Access control
- Adversarial example
- Adversarial robustness
- Adversarial suffix
- Adversarial test set
- Adversarial training
- Answer relevancy
- Attack success rate
Where you learn it
Verbosity bias is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.