Verbosity Bias
Length reads as effort. A model grading answers prefers the longer one, even when it says less.
ExampleSix padded paragraphs beat two exact sentences.
What to doWhen comparing answers, trim them to the same length or judge against a checklist, so extra words cannot win on their own.
Paste this into the chatCompare these two answers on correctness and completeness only. Length is not a merit: if the shorter one says the same thing, prefer it.