ShareGPT-gpt-oss-120B-reasoning

- ShareGPT-style multi-turn distillation with gpt-oss-120B (reasoning) teacher.
- There is a misannotation in the generation section: I used medium‑reasoning dialogues (which fit within the context window), whereas high‑reasoning multi‑turn conversations would exceed it. Originally, when annotating, I intended to use high‑reasoning ones.

本数据集为采用GPT-OSS-120B（推理）作为教师模型的类ShareGPT格式多轮蒸馏数据集。生成环节存在一处标注失误：本次实际使用的是适配上下文窗口的中等推理层级对话，而高推理层级多轮对话会超出该窗口限制。最初开展标注工作时，我本计划使用高推理层级对话样本。