wuqi28/Knowledge-Tracing-QA-Chinese
收藏资源简介:
这是一个用于知识追踪(KT)领域的高质量思维链(CoT)微调数据集,采用标准JSONL格式。每条数据包含四个核心字段:system(动态系统提示词,用于锚定模型在知识追踪、智能教育或数据挖掘领域的专家身份)、instruction(核心交互指令,涵盖知识追踪领域的模型评估方法、公式推导、算法对比等高难度专业问题)、input(补充上下文或外部输入数据,如学生答题序列,若无可为空字符串)以及output(黄金标准回复,全量包含原生推理思维链,包裹在`<think>...</think>`标签内,并输出极高质量的结构化学术级解答)。数据集旨在支持大模型在知识追踪相关任务中的微调,提升其专业问答和推理能力。
This is a high-quality Chain-of-Thought (CoT) fine-tuning dataset for the Knowledge Tracing (KT) domain, in standard JSONL format. Each data entry contains four core fields: system (dynamic system prompt to anchor the models expert identity in knowledge tracing, educational AI, or data mining), instruction (core interaction instruction covering high-difficulty professional problems in KT such as model evaluation methods, formula derivation, and algorithm comparison), input (supplementary context or external input data, e.g., student answer sequences, empty if not available), and output (gold-standard response, fully including native reasoning chain-of-thought wrapped in `<think>...</think>` tags, followed by extremely high-quality structured academic-level answers). The dataset is designed to support fine-tuning of large models for knowledge tracing-related tasks, enhancing their professional question-answering and reasoning capabilities.




