typhoon-ai/typhoon-s-instruct-post-training

该数据集是用于Typhoon-S配方的后训练语料库，旨在构建高性能、区域和领域特定的大型语言模型（LLMs），这些模型保持本地化、可控和资源高效。它采用两部分混合策略：目标语言（泰语）对齐数据和通用英语指令+工具使用数据，以保持和增强泰语本地的指令跟随、文化/语言基础以及泰英代码转换的鲁棒性，同时传递广泛有用的助手行为。数据集包括SFT分割用于监督指令调优，以及distill分割用于策略上蒸馏（OPD/GKD风格）训练。数据实例以聊天格式存储，可能包括工具定义用于工具使用/功能调用风格训练。

This dataset is a post-training corpus used in the Typhoon-S recipe for building Sovereign AI: high-performing, region- and domain-specific LLMs that remain localized, controllable, and resource-efficient. It is designed to help transform a sovereignty-adapted base model into a capable assistant while preserving target-language strengths. The dataset follows a two-part mixture philosophy: target-language (Thai) alignment data to preserve and strengthen Thai-native instruction following, cultural/linguistic grounding, and Thai–English code-switching robustness, and general English instruction + tool-use data to transfer broadly useful assistant behaviors. The dataset includes an SFT split for supervised instruction tuning, and a distill split intended for On-Policy Distillation (OPD / GKD-style) training. Examples are stored in a chat-style format and may optionally include tool definitions for tool-use / function-calling style training.