---
title: TAT-QA-Arithmetic-CoT
canonical_url: "https://www.modelscope.cn/datasets/cerebras/TAT-QA-Arithmetic-CoT"
md_url: "https://www.modelscope.cn/datasets/cerebras/TAT-QA-Arithmetic-CoT.md"
repository: cerebras/TAT-QA-Arithmetic-CoT
last_updated: 2025-10-23
license: cc-by-nc-4.0
storage_size: "25 MB"
downloads: 189
stars: 0
---

# TAT-QA-Arithmetic-CoT

> TAT-QA-Arithmetic-CoT - cerebras 在 ModelScope 开源的数据集。A Chain of Thought (CoT) version of the TAT-QA arithmetic dataset (hosted at https://huggingface.co/datasets/nvidia/ChatQA-Training-Data). The dataset was synthetically generated by prompting Llama3 70B…

cerebras/TAT-QA-Arithmetic-CoT 是 ModelScope 魔搭社区上的数据集，存储大小 25 MB，采用 cc-by-nc-4.0 许可。

- **Repository**: cerebras/TAT-QA-Arithmetic-CoT
- **License**: cc-by-nc-4.0
- **Storage size**: 25 MB
- **Downloads**: 189
- **Stars**: 0
- **Last updated**: 2025-10-23

Source: https://www.modelscope.cn/datasets/cerebras/TAT-QA-Arithmetic-CoT

---

# Dataset Information

A Chain of Thought (CoT) version of the TAT-QA arithmetic dataset (hosted at https://huggingface.co/datasets/nvidia/ChatQA-Training-Data). The dataset was synthetically generated by prompting Llama3 70B Instruct. The dataset was created as part of our work on Cerebras DocChat - a document-based conversational Q&A model. We observed that initial iterations of our model frequently made errors on arithmetic tasks (such as ConvFinQA) because it was trained on datasets such as TAT-QA where the model must create a final equation in a single shot. We found that the addition of this dataset led to a substantial boost in accuracy (+10 on ConvFinQA).

# Acknowledgement

This dataset was is a variation of the TAT-QA dataset, and was synthetically generated using Llama 3 70B Instruct.

```
@inproceedings{zhu2021tat,
  title={TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance},
  author={Zhu, Fengbin and Lei, Wenqiang and Huang, Youcheng and Wang, Chao and Zhang, Shuo and Lv, Jiancheng and Feng, Fuli and Chua, Tat-Seng},
  booktitle={Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics},
  year={2021}
}

@article{llama3modelcard,
  title={Llama 3 Model Card},
  author={AI@Meta},
  year={2024},
  url = {https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md}
}
```