Gemma 3
Google DeepMind — Gemma
rich detail 17 method fields published
Evaluation
fine-tuning of multimodal Gemma 3 pre-trained checkpoints following the Steiner et al. (2024) protocol (only learning rate swept, otherwise same transfer settings), compared with PaliGemma 2 on multimodal benchmarks; PaliGemma 2 transferred at 896x896 for the first four benchmarks and 448x448 for the others
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.
rich detail 17 method fields published