almanach/topxgen-llama-4-scout-Decomp

该数据集是用于机器翻译的合成数据集，包含源语言句子、目标语言翻译、源语言、目标语言、源句子分解得到的短语及其翻译等信息。数据集基于Llama-4-Scout-17B-16E-Instruct模型生成的中间推理轨迹，旨在训练能够进行推理的机器翻译模型。

This dataset is a synthetic dataset for machine translation, containing the source language sentence, target language translation, source language, target language, phrases obtained by decomposing the source sentence, and their translations. The dataset is based on the intermediate reasoning traces generated by the Llama-4-Scout-17B-16E-Instruct model, aiming to train reasoning-based machine translation models.