File size: 3,337 Bytes
ceadf90
 
 
 
 
 
 
 
 
 
 
 
 
 
 
84b2c6c
ceadf90
84b2c6c
 
ceadf90
 
 
 
 
09a45a5
 
 
 
 
 
 
 
ceadf90
09a45a5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
---
dataset_info:
  features:
  - name: Index
    dtype: int32
  - name: Question
    dtype: string
  - name: Options
    dtype: string
  - name: Correct Answer
    dtype: string
  - name: image
    dtype: image
  splits:
  - name: train
    num_bytes: 3145724
    num_examples: 300
  download_size: 3063257
  dataset_size: 3145724
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
task_categories:
  - visual-question-answering
tags:
  - multimodal
  - vision-language
  - clip
  - benchmark
---

# MMVP (Multimodal Visual Patterns) Benchmark

This is a corrected version of the [MMVP benchmark](https://huggingface.co/datasets/MMVP/MMVP), re-hosted by [lmms-lab-eval](https://huggingface.co/lmms-lab-eval) for use with [lmms-eval](https://github.com/EvolvingLMMs-Lab/lmms-eval).

## Why this copy?

The original `MMVP/MMVP` dataset was uploaded in `imagefolder` format, which only exposes the `image` column. The text annotations (`Question`, `Options`, `Correct Answer`, `Index`) from the accompanying `Questions.csv` were not loaded into the dataset, making it unusable for evaluation.

This version reconstructs the complete dataset by merging the images with `Questions.csv` from the original repository.

## Ground Truth Corrections

Based on verification in [lmms-eval issue #1018](https://github.com/EvolvingLMMs-Lab/lmms-eval/issues/1018) and the [original MMVP issue #30](https://github.com/tsb0601/MMVP/issues/30), we found that two pairs of samples had their answers swapped. The corrections are applied directly in this dataset:

| Index | Question | Original GT | Corrected GT | Reason |
|:-----:|:---------|:-----------:|:------------:|:-------|
| 99 | Does the elephant have long or short tusks? | (a) Long | **(b) Short** | Image shows short tusks |
| 100 | Does the elephant have long or short tusks? | (b) Short | **(a) Long** | Image shows long tusks |
| 279 | Is the elderly person standing or sitting? | (a) Standing | **(b) Sitting** | Image shows person sitting on bench |
| 280 | Is the elderly person standing or sitting? | (b) Sitting | **(a) Standing** | Image shows person standing |

## Dataset Structure

| Field | Type | Description |
|-------|------|-------------|
| `Index` | int32 | 1-based sample index (1–300) |
| `Question` | string | The visual question |
| `Options` | string | Answer choices in format `(a) ... (b) ...` |
| `Correct Answer` | string | Ground truth: `(a)` or `(b)` |
| `image` | image | 224×224 RGB image |

- **300 samples** organized as **150 pairs**
- Each pair has the same question but opposite correct answers
- Tests 9 visual patterns: orientation, direction, color, counting, etc.

## References

- **Paper**: [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://arxiv.org/abs/2401.06209)
- **Original Repository**: https://github.com/tsb0601/MMVP
- **Original Dataset**: https://huggingface.co/datasets/MMVP/MMVP
- **lmms-eval Task**: https://github.com/EvolvingLMMs-Lab/lmms-eval/tree/main/lmms_eval/tasks/mmvp

## Citation

```bibtex
@inproceedings{tong2024eyes,
  title={Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs},
  author={Tong, Shengbang and Liu, Zhuang and Zhai, Yuexiang and Ma, Yi and LeCun, Yann and Xie, Saining},
  booktitle={CVPR},
  year={2024}
}
```