CMATH

Overview

CMATH is a Chinese elementary school mathematics benchmark containing 1,698 problems across grades 1-6. It evaluates the mathematical reasoning capabilities of language models on Chinese-language math word problems at increasing difficulty levels.

Task Description

Task Type: Chinese Mathematical Word Problem Solving
Input: Chinese math word problem (elementary school level)
Output: Step-by-step reasoning with numerical answer
Difficulty: Grade 1 (easiest) to Grade 6 (hardest)

Key Features

1,098 test problems + 600 validation problems
Six grade levels (1-6) for fine-grained difficulty analysis
Problems are in Chinese, testing language-specific reasoning
Simple numerical answers (integers or decimals)
Metadata includes reasoning steps count and digit complexity

Evaluation Notes

Default configuration uses 0-shot evaluation
Answers should be formatted within \boxed{} for proper extraction
Numeric accuracy metric for answer comparison
Results can be broken down by grade level
Chinese prompt template used by default

Properties

Property	Value
Benchmark Name	`cmath`
Dataset ID	evalscope/cmath
Paper	N/A
Tags	`Chinese`, `Math`, `Reasoning`
Metrics	`acc`
Default Shots	0-shot
Evaluation Split	`test`

Data Statistics

Metric	Value
Total Samples	1,098
Prompt Length (Mean)	70.14 chars
Prompt Length (Min/Max)	38 / 191 chars

Per-Subset Statistics:

Subset	Samples	Prompt Mean	Prompt Min	Prompt Max
`Grade 1`	164	60.65	45	99
`Grade 2`	253	63.3	38	168
`Grade 3`	237	69.49	49	148
`Grade 4`	120	75.33	51	114
`Grade 5`	126	76.79	49	124
`Grade 6`	198	80.16	42	191

Sample Example

Subset: Grade 1

{
  "input": [
    {
      "id": "6e98e414",
      "content": "妈咪买了3盒茶叶，一盒茶叶有6小包，一共买了多少小包茶叶？\n请一步一步推理，最后将答案放在\\boxed{}中。\n"
    }
  ],
  "target": "18",
  "id": 0,
  "group_id": 0,
  "subset_key": "Grade 1",
  "metadata": {
    "reasoning_step": 1,
    "num_digits": 2
  }
}

Prompt Template

Prompt Template:

{question}
请一步一步推理，最后将答案放在\boxed{{}}中。

Usage

Using CLI

evalscope eval \
    --model YOUR_MODEL \
    --api-url OPENAI_API_COMPAT_URL \
    --api-key EMPTY_TOKEN \
    --datasets cmath \
    --limit 10  # Remove this line for formal evaluation

Using Python

from evalscope import run_task
from evalscope.config import TaskConfig

task_cfg = TaskConfig(
    model='YOUR_MODEL',
    api_url='OPENAI_API_COMPAT_URL',
    api_key='EMPTY_TOKEN',
    datasets=['cmath'],
    dataset_args={
        'cmath': {
            # subset_list: ['Grade 1', 'Grade 2', 'Grade 3']  # optional, evaluate specific subsets
        }
    },
    limit=10,  # Remove this line for formal evaluation
)

run_task(task_cfg=task_cfg)

3.2 KiB Raw Blame History Unescape Escape