oliverkinch/da-instruct-dynaword-contemporary-hq

这是一个通过反向翻译从danish-foundation-models/danish-dynaword生成的丹麦语指令微调数据集，并使用danish-foundation-models/dynaword-annotations进行高质量当代丹麦语样本的过滤。排除了主要是历史或古丹麦语的子集。每一行数据都是一个(prompt, target)对，其中target是来自DynaWord精选子集的真实丹麦语文本段落，prompt是通过Qwen3.5-397B-A17B模型使用反向翻译方法生成的现实丹麦语用户指令，这些指令会引发语言模型产生类似的文本。数据集包含11,303个指令对，来自28个DynaWord子集的9,200个随机抽样源ID。

Danish instruction fine-tuning dataset generated via backtranslation from danish-foundation-models/danish-dynaword, filtered to high-quality contemporary Danish samples using danish-foundation-models/dynaword-annotations. Subsets consisting primarily of historical or archaic Danish are excluded. Each row is a (prompt, target) pair where target is a passage of authentic Danish text drawn from a curated subset of DynaWord, and prompt is a realistic Danish user instruction generated by Qwen3.5-397B-A17B using a backtranslation approach that would plausibly elicit that text from a language model. The dataset contains 11,303 instruction pairs generated from 9,200 randomly sampled source IDs across 28 dynaword subsets.