---
language: vi
tags:
  - math
  - llm
  - fine-tuning
license: apache-2.0
datasets:
  - microsoft/orca-math-word-problems-200k
model-index:
  - name: My-Math-LLM
    results:
      - task:
          type: text-generation
        dataset:
          name: microsoft/orca-math-word-problems-200k
          type: text
        metrics:
          - type: ROUGE_1, ROUGE_2, ROUGE_L
            value: 0.5578, 0.3020, 0.5114
---

# Model Card for Math LLM (Vietnamese)

## Model Details

### Model Description

This model is a fine-tuned version of **VinaLLaMA-7B-Chat**, trained to solve elementary to secondary school math problems in Vietnamese. It utilizes the **Microsoft ORCA Math Word Problems 200k** dataset, enabling it to understand and solve word problems with high accuracy.

- **Developed by:** [@Namronaldo2024](https://huggingface.co/Namronaldo2004), [@Kiet2302](https://huggingface.co/Kiet2302), 
- **Model type:** Large Language Model (LLM) for Math Question-Answering
- **Language(s) (NLP):** Vietnamese
- **License:** Apache 2.0
- **Finetuned from model:** [vilm/vinallama-7b-chat](https://huggingface.co/vilm/vinallama-7b-chat)

### Model Sources

- **Repository:** [Source Github](https://github.com/thiennbao/zundamath)

## Uses

### Direct Use

This model can be used directly to solve elementary/secondary level math problems.

### Downstream Use

This model can be fine-tuned on other math dataset to improve the context understanding in other fields such as physics or chemistry.

### Out-of-Scope Use

This model is designed not to:
- Answer questions that are out of math scope.
- Deeply understand the complex/high level math theory.

## Bias, Risks, and Limitations

This model can cause some errors when:
- Solve difficult math problems that have different ways to solve.
- Math problems with ambiguous description.

### Recommendations

Users should not use the result from the bot without validation.