Gemma 1
Google DeepMind — Gemma
rich detail 23 method fields published
Evaluation
Automated Benchmarks
performance across physical reasoning, social reasoning, question answering, coding, mathematics, commonsense reasoning, language modeling, and reading comprehension; evaluation methodology largely shared with Gemini and replicated from the Mistral technical report for ARC, CommonsenseQA, Big Bench Hard and AGI Eval
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.