13274243a0
Keep K3 suite selection and report-schema scoring in bash, merge K3/vision dataset_args into dpv4 yamls, and pin EvalScope at 735d920ee911 with local patches. Co-authored-by: Cursor <cursoragent@cursor.com>
95 lines
2.5 KiB
Markdown
95 lines
2.5 KiB
Markdown
# General-VMCQ
|
|
|
|
|
|
## Overview
|
|
|
|
General-VMCQ is a customizable visual multiple-choice question answering benchmark for multimodal models.
|
|
It uses MMMU-style format with image/video/audio placeholders in text, supporting flexible media inputs.
|
|
|
|
## Task Description
|
|
|
|
- **Task Type**: Visual Multiple-Choice Question Answering
|
|
- **Input**: Question with `<image N>`/`<video N>`/`<audio N>` placeholders + choice options + media
|
|
- **Output**: Selected answer choice
|
|
- **Flexibility**: Supports custom datasets via local files
|
|
|
|
## Key Features
|
|
|
|
- MMMU-style format (not OpenAI message format), supporting up to 100 images/videos/audios per sample
|
|
- Flexible image/video/audio input (path, URL, base64 data URL, or Hugging Face media columns with `{"path": ...}` or `{"bytes": ...}`)
|
|
- Additional format support for `images`/`videos`/`audios` column for unlimited list of media
|
|
- Chain-of-thought prompt template option
|
|
- Custom dataset support via local file loading
|
|
|
|
## Evaluation Notes
|
|
|
|
- Default configuration uses **0-shot** evaluation
|
|
- Primary metric: **Accuracy**
|
|
- Train split: **dev**, Eval split: **val**
|
|
- Raw media bytes require parquet; plaintext files carry paths, URLs, or base64 data URLs
|
|
- See [User Guide](https://evalscope.readthedocs.io/en/latest/advanced_guides/custom_dataset/vlm.html) for dataset format
|
|
|
|
|
|
## Properties
|
|
|
|
| Property | Value |
|
|
|----------|-------|
|
|
| **Benchmark Name** | `general_vmcq` |
|
|
| **Dataset ID** | `general_vmcq` |
|
|
| **Paper** | N/A |
|
|
| **Tags** | `Custom`, `MCQ`, `MultiModal` |
|
|
| **Metrics** | `accuracy` |
|
|
| **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:**
|
|
```text
|
|
Answer the following multiple choice question. The last line of your response should be of the following format: 'ANSWER: [LETTER]' (without quotes) where [LETTER] is one of {letters}. Think step by step before answering.
|
|
|
|
{question}
|
|
|
|
{choices}
|
|
```
|
|
|
|
## Usage
|
|
|
|
### Using CLI
|
|
|
|
```bash
|
|
evalscope eval \
|
|
--model YOUR_MODEL \
|
|
--api-url OPENAI_API_COMPAT_URL \
|
|
--api-key EMPTY_TOKEN \
|
|
--datasets general_vmcq \
|
|
--limit 10 # Remove this line for formal evaluation
|
|
```
|
|
|
|
### Using Python
|
|
|
|
```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_vmcq'],
|
|
limit=10, # Remove this line for formal evaluation
|
|
)
|
|
|
|
run_task(task_cfg=task_cfg)
|
|
```
|