This dataset contains prompt-only rows extracted from a-m-team/AM-DeepSeek-R1-Distilled-1.4M.
For each source JSONL row, every message with role == "user" was emitted as one prompt row. Assistant responses, reasoning traces, answers, and source metadata were not included.
Source files:
Input rows: 1,400,000
Output prompt rows: 1,400,000
JSON parse errors: 0… See the full description on the dataset page:
https://huggingface.co/datasets/tim-gabie/am-deepseek-r1-distilled-prompts-1.4m.
