Losa10/G3P-Finetuning-examples

一个高质量的合成数据集，专为**指令微调**和**推理（CoT）**开发而设计。该数据集使用**Gemini 3 Pro**预览模型生成，专注于技术任务、复杂配置和逻辑分步问题解决。数据集包含两种配置：一种是包含完整推理链（Chain-of-Thought）的复杂任务，另一种是用于直接响应训练的标准指令-输出对。数据集支持英语和俄语，适用于监督微调（SFT）、推理能力训练和多语言对齐等应用场景。

A high-quality synthetic dataset designed for **Instruction Fine-Tuning** and **Reasoning (CoT)** development. Generated using the **Gemini 3 Pro** preview model, this dataset focuses on technical tasks, complex configurations, and logical step-by-step problem-solving. The dataset is split into two distinct configurations: one includes full reasoning chains (Chain-of-Thought) for complex tasks, and the other provides standard instruction-output pairs for direct response training. It supports both English and Russian languages and is intended for use cases such as Supervised Fine-Tuning (SFT), Reasoning Capability enhancement, and Multilingual Alignment.