Gemma: Open Models Based on Gemini Research and Technology

Citations Over Time

Abstract

This work introduces Gemma, a family of lightweight, state-of-the art open models built from the research and technology used to create Gemini models. Gemma models demonstrate strong performance across academic benchmarks for language understanding, reasoning, and safety. We release two sizes of models (2 billion and 7 billion parameters), and provide both pretrained and fine-tuned checkpoints. Gemma outperforms similarly sized open models on 11 out of 18 text-based tasks, and we present comprehensive evaluations of safety and responsibility aspects of the models, alongside a detailed description of model development. We believe the responsible release of LLMs is critical for improving the safety of frontier models, and for enabling the next wave of LLM innovations.

Related Papers

• → Control of vegetative reproduction in Marchantia polymorpha by the KAI2-ligand signaling pathway(2023)24 cited
• → CRYO SEM EXAMINATION OF GEMMA FORMATION IN VITTARIA GRAMINIFOLIA(1988)13 cited
• → Growth, Development and Gemma Formation in the Liverwort Scapania nemorosa as Influenced by L-Arginine, L-Histidine and L-Glutamic Acid(1965)3 cited
• Susquehanna Chorale Spring Concert "Roots and Wings"(2017)
• → KARRIKIN INSENSITIVE2 ( KAI2 )-dependent signaling pathway controls vegetative reproduction in Marchantia polymorpha(2022)