Question about results

#24
by arijit-kensho - opened

Are the results on various benchmarks generated using the instruct models or the base models?

Google org

Hi, the academic benchmarks are based on the base models while the safety evals are on the it models. More details on benchmarking is in the tech report here: https://storage.googleapis.com/deepmind-media/gemma/gemma-report.pdf

manvinder changed discussion status to closed

Sign up or log in to comment