Claude-Opus-4.6-Reasoning-927x

# Claude Opus 4.6 - High Reasoning

This is a reasoning dataset generated using Claude Opus 4.6 with high reasoning effort

It contains distilled reasoning traces from Bullshit Bench for bullshit detection, legal and life decisions data for generalization, traces for improving the models understanding of vague and lazy prompts and more.

## Formatting guide

```json
{
"messages": [
{"role": "user", "content": "..."},
{"role": "assistant", "thinking": "...", "content": "Final answer..."}
]
}
```

`thinking` is only included when a reasoning trace is present. All other fields are preserved as-is.

## Stats

- **Cost:** $ 49.70 (USD)
- **Tokens (input + output):** 2.04 M

---

This dataset was generated using [**TeichAI/datagen**](https://github.com/TeichAI/datagen). Check it out to see how this dataset was made and/or to make datasets like these.

# Claude Opus 4.6——高推理能力数据集
本数据集为采用Claude Opus 4.6模型，以高推理算力生成的推理类数据集。
其包含源自废话检测基准（Bullshit Bench）的精炼推理轨迹、用于提升模型泛化能力的法律与生活决策数据集，以及用于提升模型对模糊及敷衍式提示词理解能力的推理轨迹及其他相关内容。

## 格式规范
json
{
"messages": [
{"role": "user", "content": "..."},
{"role": "assistant", "thinking": "...", "content": "最终答案..."}
]
}

仅当存在推理轨迹时，才需包含`thinking`字段，其余字段均需保留原始格式。

## 统计信息
- **成本：** 49.70美元（USD）
- **Token（输入+输出）总量：** 2.04 M

---
本数据集通过[**TeichAI/datagen**](https://github.com/TeichAI/datagen)生成，可访问该仓库了解本数据集的生成流程，或自行创建同类数据集。