Gemma 3
Google DeepMind — Gemma
partial detail 15 method fields published
Evaluation
Sliding window size ablation (Gemma 3)
sliding window sizes for local attention layers across different global:local ratio configurations (two 2B models, 1:1 and 1:3 local:global ratios; text-only)
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.