Skip to content
Trending storyDeveloping

Artificial Analysis splits model rubric scores into Climb and Regression criteria

1 article1 sourcesince Nov 10Last article 1h ago ·

Overview

AISummary of 1 article

Artificial Analysis says it labels each rubric criterion as either Climb or Regression.

Climb criteria test harder tasks where frontier models still often fail. Regression criteria test baseline competence.

On Regression criteria, the leading seven models all score between 95% and 97%, so the scores barely separate them. Climb criteria separate the models more widely, with scores from 28% to 57%.

Written by AI from the articles below · updated Nov 10, 1:29 PM ET

Check the sources:

Latest developmentNov 10, 1:16 PM ET
Artificial Analysis reports model scores on Climb and Regression rubric criteria

Article timeline

The articles in this story. Times are ET.

Nov 10
  1. Artificial AnalysisOfficial
    Artificial Analysis reports model scores on Climb and Regression rubric criteria

    AIArtificial Analysis says each rubric criterion is labeled Climb, for harder criteria where frontier models still often fail, or Regression, for baseline competence. On Regression, the leading seven models all score between 95% and 97%, so there is little separation at the top. Climb separates the models more, with scores ranging from 28% to 57%.

    Image from @ArtificialAnlys's post

Heat Trend

Not enough continuous observations to show a trend yet.