This dataset contains 7,680 audited completions from the official
post-trained Qwen/Qwen3-8B model at revision
b968826d9c46dd6066d109eabc6255188de91218. It covers AIME I and AIME II from 2009 through
2024: exactly 30 problems per year, 480 unique prompts, and 16 independently
seeded completions per prompt.
Thinking is enabled through Qwen's chat template (enable_thinking=True).
There… See the full description on the dataset page:
https://huggingface.co/datasets/suryadv/qwen3-8b-aime-2009-2024-16x.