tulu-3-sft-olmo-2-mixture-0225

Used to train OLMo 2 32B. From the [blog post](https://allenai.org/blog/olmo2-32B):
> Filtered out instructions from the SFT dataset and the chosen responses of the preference data that included mentions of a date cutoff from the synthetic data generation process. This resulted in a new version of the instruction dataset, Tulu 3 SFT Mixture 0225, and preference dataset, OLMo-2-32B-pref-mix-0325.
> We use majority voting to improve the quality of answers to our synthetic math questions. For our Persona MATH and Grade School Math datasets from Tülu 3, we only include prompts and completions where the model reaches a majority vote over 5 completions. New versions of the math and grade school math datasets are available.

Created with `open-instruct` data tools:
```
python scripts/data/filtering_and_updates/update_subsets.py \
--base_ds allenai/tulu-3-sft-olmo-2-mixture-filter-datecutoff \
--remove_sources ai2-adapt-dev/personahub_math_v5_regen_149960 allenai/tulu-3-sft-personas-math-grade \
--add_ds allenai/tulu-3-sft-personas-math-filtered allenai/tulu-3-sft-personas-math-grade-filtered \
--remove_keys prompt dataset \
--push_to_hub \
--repo_id allenai/tulu-3-sft-olmo-2-mixture-0225
```

本数据集用于训练OLMo 2 32B模型，相关详情可参阅[博客文章](https://allenai.org/blog/olmo2-32B)：
> 我们从监督微调（Supervised Fine-Tuning，SFT）数据集以及偏好数据的已选回复中，剔除了所有提及合成数据生成流程中日期截断相关内容的指令条目。本次处理生成了两个新版本数据集：指令数据集Tulu 3 SFT Mixture 0225，以及偏好数据集OLMo-2-32B-pref-mix-0325。
> 针对合成数学问题的回答质量，我们采用多数投票机制进行优化；对于Tulu 3中的Persona MATH与Grade School Math数据集，我们仅保留模型在5次回复中达成多数共识的提示词与回复内容。新版数学及中小学数学数据集现已发布。

本数据集通过`open-instruct`数据工具构建，执行的处理命令如下：

python scripts/data/filtering_and_updates/update_subsets.py
--base_ds allenai/tulu-3-sft-olmo-2-mixture-filter-datecutoff
--remove_sources ai2-adapt-dev/personahub_math_v5_regen_149960 allenai/tulu-3-sft-personas-math-grade
--add_ds allenai/tulu-3-sft-personas-math-filtered allenai/tulu-3-sft-personas-math-grade-filtered
--remove_keys prompt dataset
--push_to_hub
--repo_id allenai/tulu-3-sft-olmo-2-mixture-0225