QwQ-LongCoT-Verified-130K

Original Dataset: [amphora/QwQ-LongCoT-130K](https://huggingface.co/datasets/amphora/QwQ-LongCoT-130K)

QwQ 32B Preview isn't perfect :)

**Note**: Around 5-7% of the processed data might be incorrectly labeled as "unverified" because QwQ's output isn't *exactly* the same as the original solution from NuminaMathCoT. I believe this can be solved with another round of processing with a smarter model but Qwen 2.5 3B Instruct is good enough to check if the solution is exactly the same. Magpie data is also "unverified" and has an empty "solution" column.

| Subset | Information | Rows |
|---------------------|---------------------------------------------------------|-----------|
| **default** | Includes all the data from the original dataset. | 133k |
| **magpie** | Just the Magpie generated data. | 43k |
| **numina-math-only**| Only the problems coming from the NuminaMathCoT dataset.| 90.1k |
| **verified** | Keeps only the verified data. | 64.6k |

原始数据集：[amphora/QwQ-LongCoT-130K](https://huggingface.co/datasets/amphora/QwQ-LongCoT-130K)

QwQ 32B 预览版并非完美无缺 :)

**注**：由于QwQ的输出与NuminaMathCoT的原始解题方案并非完全一致，约5%至7%的已处理数据可能被错误标记为“未验证”。笔者认为可通过使用更智能的模型开展第二轮处理以解决该问题，但Qwen 2.5 3B Instruct已足够用于校验解题方案是否完全匹配。Magpie数据同样处于“未验证”状态，且其"solution"（解题方案）字段为空。

| 子集名称 | 信息说明 | 数据行数 |
|---------------------|---------------------------------------------------------|-----------|
| **default** | 包含原始数据集的全部数据。 | 13.3万 |
| **magpie** | 仅包含Magpie生成的数据。 | 4.3万 |
| **numina-math-only**| 仅包含源自NuminaMathCoT数据集的题目。| 9.01万 |
| **verified** | 仅保留已验证的数据。 | 6.46万 |

搜集汇总

背景与挑战

背景概述

该数据集基于原始数据集QwQ-LongCoT-130K构建，包含约133k行数据，并划分为多个子集，其中已验证的子集有64.6k行。部分数据可能因模型输出与原始解决方案不完全一致而被错误标记为未验证。