General-MCQ

Overview

General-MCQ is a customizable multiple-choice question answering benchmark for evaluating language models. It supports flexible data formats and variable number of answer choices.

Task Description

Task Type: Multiple-Choice Question Answering
Input: Question with 2-10 answer choices (A through J)
Output: Selected answer choice(s)
Flexibility: Supports custom datasets via local files, single or multiple correct answers

Key Features

Flexible number of choices (A through J)
Custom dataset support via local file loading
Chinese single-/multiple-answer prompt templates (optional CoT variants)
Configurable few-shot examples
Accuracy-based evaluation

Evaluation Notes

Default configuration uses 0-shot evaluation with single-answer template
Set extra_params.multiple_correct=True to evaluate questions with multiple correct answers
Set extra_params.use_cot=True to switch to chain-of-thought prompt templates
Primary metric: Accuracy
Train split: dev, Eval split: val
See User Guide for dataset format

Properties

Property	Value
Benchmark Name	`general_mcq`
Dataset ID	`general_mcq`
Paper	N/A
Tags	`Custom`, `MCQ`
Metrics	`acc`
Default Shots	0-shot
Evaluation Split	`val`
Train Split	`dev`

Data Statistics

Statistics not available.

Sample Example

Sample example not available.

Prompt Template

Prompt Template:

回答下面的单项选择题，请选出其中的正确答案。你的回答的全部内容应该是这样的格式："答案：[LETTER]"（不带引号），其中 [LETTER] 是 {letters} 中的一个。

问题：{question}
选项：
{choices}

Extra Parameters

Parameter	Type	Default	Description
`multiple_correct`	`bool`	`False`	Whether the dataset contains questions with multiple correct answers. When True, switches to the multiple-answer prompt template and parser, and requires the `answer` field to be a list of letters (e.g., ["A", "C"]).
`use_cot`	`bool`	`False`	Whether to use the chain-of-thought (CoT) prompt template variant.

Usage

Using CLI

evalscope eval \
    --model YOUR_MODEL \
    --api-url OPENAI_API_COMPAT_URL \
    --api-key EMPTY_TOKEN \
    --datasets general_mcq \
    --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=['general_mcq'],
    dataset_args={
        'general_mcq': {
            # extra_params: {}  # uses default extra parameters
        }
    },
    limit=10,  # Remove this line for formal evaluation
)

run_task(task_cfg=task_cfg)

3.0 KiB Raw Blame History