Skip to content
Trending storyDeveloping

Tazo Labs launches DECIDE, a benchmark for voice-agent decisions on live calls

1 article1 sourcesince Nov 7Last article 3h ago ·

Overview

AISummary of 1 article

Tazo Labs introduces DECIDE, a benchmark that tests voice agents on decisions made during real production calls, such as when to speak, route a call, pick a tool, or switch language.

The benchmark contains 4,867 decision questions drawn from 1,081 real agent-caller clips of 3 to 25 seconds, spanning 32 tasks in six families.

DECIDE scores 20 models, including direct-audio and speech-to-text plus text models, on the same questions. Tazo Labs says any model endpoint can be submitted for scoring without an invite.

Written by AI from the articles below · updated Nov 7, 12:08 PM ET

Check the sources:

Latest developmentNov 7, 9:27 AM ET
Tazo Labs introduces DECIDE, a benchmark for voice agent decisions

Article timeline

The articles in this story. Times are ET.

Nov 7
  1. bolnaX
    Tazo Labs introduces DECIDE, a benchmark for voice agent decisions

    AITazo Labs introduces DECIDE, a benchmark that evaluates voice agents on the decisions they make during real production calls, such as when to speak, route a call, pick a tool, or switch language. DECIDE contains 4,867 decision questions drawn from 1,081 real agent-caller clips of 3 to 25 seconds, across 32 tasks in six families. It scores 20 direct-audio and speech-to-text plus text models on the same questions, and Tazo Labs says any model endpoint can be submitted for scoring without an invite.

Heat Trend

Not enough continuous observations to show a trend yet.