almanach/topxgen-llama-4-scout-CoT

该数据集是一个用于机器翻译的合成数据集，它通过在大型语言模型中引入中间推理标记来增强翻译结果。每个数据条目包含源语言句子、目标翻译、源语言和目标语言信息，以及用于CoT微调的推理标记和翻译的连接序列。该数据集支持Xhosa语言，并可用于微调LLM进行基于推理的翻译、研究MT中的思考标记以及比较CoT微调与直接输出微调在MT设置中的效果。

This dataset is a synthetic dataset for machine translation that enhances translation results by introducing intermediate reasoning tokens in large language models. Each entry contains a source language sentence, target translation, source and target language information, and a concatenated sequence of reasoning tokens and translation for CoT fine-tuning. The dataset supports the Xhosa language and can be used for fine-tuning LLMs for reasoning-based translation, studying thinking tokens in MT, and comparing CoT fine-tuning with direct output fine-tuning in MT setups.