---
dataset_info:
  features:
  - name: ID
    dtype: string
  - name: Type
    dtype: string
  - name: Education_Level
    dtype: string
  - name: Grade
    dtype: string
  - name: Question
    dtype: string
  - name: Answer
    dtype: string
  - name: Explanation
    dtype: string
  - name: Difficulty_Level
    dtype: string
  - name: Categories
    dtype: string
  - name: Source
    dtype: string
  splits:
  - name: Elementary
    num_bytes: 85608
    num_examples: 332
  - name: Middle
    num_bytes: 1001357
    num_examples: 1108
  - name: High
    num_bytes: 154761
    num_examples: 426
  download_size: 550840
  dataset_size: 1241726
configs:
- config_name: default
  data_files:
  - split: Elementary
    path: data/Elementary-*
  - split: Middle
    path: data/Middle-*
  - split: High
    path: data/High-*
license: apache-2.0
---

# Model Performance Overview

## Base Models

| Model Name   | Parameters | Inference Accuracy | Fine-tune Accuracy  |Notes                                                           |
| ------------ | ---------- | ------------------ | ------------------  |--------------------------------------------------------------- |
| Qwen 2.5 7B  | 7B         | 64.68              | 68.7               |[Model Link](https://huggingface.co/DHMATH/Qwen_2.5-Math-v3)    |                                                           |
| Mistral v0.3 | 7B         | 30.44              | -                  | -                                                              |                                                            |

## Math-Specialized Models

| Model Name    | Parameters | Inference Accuracy  | Notes                                                           |
| ------------- | ---------- | ------------------  | --------------------------------------------------------------- |
| Qwen2.5-Math  | 7B         | 54.34              | -                                                               |
| Deepseek-Math | 7B         | 52.3               | -                                                               |
| Mathstral     | 7B         | 26.37             | -                                                               |
| MAmmoTH-Plus  | 7B         | 43.14               | [Model Link](https://huggingface.co/TIGER-Lab/MAmmoTH2-7B-Plus) |

## Proprietary Models

| Model Name | Parameters | Inference Accuracy | 
| ---------- | ---------- | ------------------ | 
| Gemini Pro | -          | 81.03              |