README.md

# TAT-LLM: A Specialized Language Model for Discrete Reasoning over Tabular and Textual Data

Paper: https://arxiv.org/abs/2401.13223

Code: https://github.com/fengbinzhu/TAT-LLM


## Introduction

We present TAT-LLM, a specialized language model crafted through the innovative Step-wise Pipeline approach, focusing on the nuanced realm of tabular and textual question answering (QA). This model is the fruit of rigorously fine-tuning the LLaMA 2 architecture with a novel dataset, autonomously generated from expertly annotated resources. TAT-LLM stands at the intersection of tabular comprehension and textual analysis, engineered to excel by embodying three fundamental phases: Extraction, Reasoning, and Execution. Our empirical findings illuminate TAT-LLM's remarkable capability to eclipse traditional benchmarks, surmounting even the most advanced models and colossal language models such as GPT-4 across a suite of demanding QA tasks like FinQA, TAT-QA, and TAT-DQA. This endeavor not only sets a new standard for task-specific language models but also paves the way for future explorations in optimizing smaller models for highly specialized functions.

| Model | Size | FINQA | TATQA | TATDQA |
| ---   | ---  | ---   | ---   | ---    |
| GPT-3.5-Turbo | - | 58.00 | 59.47 | 52.74 |
| GPT-4 | - | 63.91 | 71.92 | 64.46 |
| TAT-LLM-7B | 7B | 65.13 | 76.49 | 71.38 |
| TAT-LLM-13B | 13B | 71.93 | 77.51 | 72.22 |
| TAT-LLM-70B | 70B | **76.81** | **81.42** | **76.55** |


## Training

We train our TAT-LLM model in various sizes, including 7B, 13B, and 70B, by fine-tuning LLaMA 2 using Low-Rank Adaptation (LoRa) on a combination of the train sets from FinQA, TAT-QA and TAT-DQA datasets. To refine accuracy, we introduce an External Executor, enhancing the model by processing intermediate outputs to derive conclusive answers. Please refer to the [paper](https://arxiv.org/abs/2401.13223) for more details.

## Inference & Evaluation

Please refer to code [here](https://github.com/fengbinzhu/TAT-LLM)

## Citation

If you find this repository helpful, please consider citing our paper:

```
@misc{zhu2024tatllm,
      title={TAT-LLM: A Specialized Language Model for Discrete Reasoning over Tabular and Textual Data},
      author={Fengbin Zhu and Ziyang Liu and Fuli Feng and Chao Wang and Moxin Li and Tat-Seng Chua},
      year={2024},
      eprint={2401.13223},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```