Skip to content

Evaluation

Impact of formatting ablation

performance variance on MMLU across 12 prompt/evaluation formatting variations as a proxy for undesired performance variability

Other documented tests Google DeepMind — Gemma

How it was run

Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.

Gemma 2
Google DeepMind — Gemma

rich detail 22 method fields published

See how this test was run