一句话：面向 Google 内部软件工程定制 Gemini，经万亿 token 数据中训练与全栈调优，显著提升编码助手效率与代码留存率。
企业软件开发是持续演进过程，产生大量可供 LLM 微调的数据。本文提出 Gemini for Google（GfG），针对 Google 内部软件工程生态改造 Gemini，覆盖从整理一万亿 token 专有数据集、设计缓解灾难性遗忘的 mid-training 策略，到继续预训练与后训练的全栈调优及下游应用部署。在覆盖 29,000 名开发者的大规模盲测 A/B 研究中，GfG 显著优于基线：每轮平均迭代次数减少 23%，AI 生成代码在 14 天后仍未被修改的留存率提升约 17%。论文进一步给出企业级模型适配蓝图，包括高价值信号抽取、数据准备、全栈调优与部署，为其他组织利用内部工程数据提供框架。
Enterprise software development is a continuous evolutionary process, characterized by incremental additions, architectural revisions, production deployments and rigorous maintenance. These activities generate valuable data that modern LLMs could be finetuned on, to unlock additional tool possibilities for enterprise software engineering. While frontier LLMs are already very capable, this form of customization offers a compelling path for enterprise-specific optimization. We introduce Gemini for Google (GfG), an adaptation of Gemini specialized for Google's internal software engineering ecosystem. This paper details the model's end-to-end development, from curating a trillion-token proprietary dataset to implementing a mid-training strategy that mitigates catastrophic forgetting. In a large-scale blind A/B study across 29,000 developers, Gemini for Google significantly outperformed baselines: reducing the mean number of iterations per turn by 23%, and increasing the percentage of AI-generated code lines that remains in the codebase unmodified after 14 days (code survival rate) by about 17%. Beyond metrics, we provide a comprehensive blueprint for enterprise model adaptation, covering: (1) the extraction of high-value signals from software engineering data, (2) data preparation strategies, (3) full-stack model tuning (continued pre-training and post-training), and (4) the deployment of downstream applications. We believe this methodology offers a guiding framework for other organizations to unlock the full potential of their internal engineering data.
@inproceedings{abs-2605-16517,
title = {Customizing an LLM for Enterprise Software Engineering},
author = {Aditya Kini and Satish Chandra and Milad Hashemi and Saksham Thakur and Aditya Pandey and Vincent Nguyen and Marc Brockschmidt and Franjo Ivančić and Danny Tarlow and Parthasarathy Ranganathan and Petros Maniatis and Ahmed Omran and Zaheer Abbas and Anita Gergely and Martin Sevenich and Gufeng Zhang and Amy Hua and Alexander Frömmgen},
booktitle = {ASE},
year = {2026},
doi = {10.1145/3832783.3834508}
}