AI Change Tracker
3

LMArena tightens style-control methodology after gaming concerns

2026-08-15

LMArena rolled out a revised style-control adjustment and de-listed several anonymous checkpoints after evidence that verbose, heavily formatted answers were inflating Elo. Rankings for current frontier models shifted by up to 12 points.

Significance 3: Changes how a trusted human-preference source should be read; no single model moved decisively.

Benchmark result Tooling / agents

Sources

Affected axes

Human preference

← Back to feed