Decoder. plain-English AI glossary

Majority voting

● Core

Asking multiple models or multiple runs of the same model and going with the most common answer.

Think of it like

A jury voting on guilt — whichever verdict gets the most votes wins.

Example

You ask five different models the same question. Three say yes, two say no. You take "yes" as the answer.

How it actually works

Majority voting works because diverse models often fail in different ways. But it's expensive — you pay for five models instead of one. It also assumes the models are somewhat independent; if they're all fine-tuned on the same data, their errors correlate and voting doesn't help. Edge case: tie votes.

For product teams

Valuable for high-stakes decisions but has a cost — use only where accuracy matters enough to pay the price.

For engineers

Collect outputs from k models, count votes, return mode. Handle ties by picking alphabetically or using a tiebreaker model.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome