beyoru/deepseek-v4-pro-max-distillation-preview-shot
收藏资源简介:
该数据集包含由DeepSeek-V4-Pro(`reasoning_effort=max`,`thinking.enabled=true`)生成的推理痕迹和最终答案,使用的提示来自[`Jackrong/GLM-5.1-Reasoning-1M-Cleaned`](https://huggingface.co/datasets/Jackrong/GLM-5.1-Reasoning-1M-Cleaned)。目标是检查质量。数据集包含500个样本,提示来源为`Jackrong/GLM-5.1-Reasoning-1M-Cleaned`的`train`分割(前500行,流式传输)。教师模型为`deepseek-v4-pro`,推理努力为`max`,语言主要为英语,包含一些中文/多语言STEM内容,格式为JSON Lines(`output.jsonl`)。每个JSON对象包含`id`、`domain`、`prompt`、`reasoning`、`response`、`model`和`usage`等字段。
This dataset contains reasoning traces and final answers generated by DeepSeek-V4-Pro (`reasoning_effort=max`, `thinking.enabled=true`) using prompts sampled from [`Jackrong/GLM-5.1-Reasoning-1M-Cleaned`](https://huggingface.co/datasets/Jackrong/GLM-5.1-Reasoning-1M-Cleaned). The goal is to check quality. The dataset contains 500 samples, with prompts sourced from the `train` split of `Jackrong/GLM-5.1-Reasoning-1M-Cleaned` (first 500 rows, streaming). The teacher model is `deepseek-v4-pro`, reasoning effort is `max`, languages are primarily English with some Chinese/multilingual STEM content, and the format is JSON Lines (`output.jsonl`). Each JSON object includes fields such as `id`, `domain`, `prompt`, `reasoning`, `response`, `model`, and `usage`.




