SUSTech/Mars-VL-Pairs

---
language:
- en
license: cc-by-4.0
size_categories:
- 1K<n<10K
task_categories:
- image-to-text
- text-to-image
dataset_info:
features:
- name: key
dtype: string
- name: image_url
dtype: string
- name: ori_caption
dtype: string
- name: refined_caption
dtype: string
splits:
- name: train
num_bytes: 893082
num_examples: 2287
download_size: 485670
dataset_size: 893082
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
tags:
- Multimodal
- Retrieval
- planet
---

# Paired Image–Text Retrieval

[**Paper**](https://huggingface.co/papers/2602.13961) | [**Code**](https://github.com/ml-stat-Sustech/MarsRetrieval)

## Dataset Summary

This dataset is Task 1 of [**MarsRetrieval**](https://github.com/ml-stat-Sustech/MarsRetrieval), a retrieval-centric benchmark for evaluating vision-language models (VLMs) on Mars geospatial discovery. Task 1 evaluates **fine-grained image–text alignment** using curated Martian image–text pairs spanning multiple spatial scales, from global orbital mosaics to rover-level imagery. The dataset contains **2,287** paired image–text samples. For details about dataset construction and evaluation protocol, please refer to the [official repository](https://github.com/ml-stat-Sustech/MarsRetrieval/blob/main/docs/DATASET.md).

## Task Formulation

We formulate this task as a **bidirectional one-to-one retrieval** benchmark:

- **Text → Image** retrieval
- **Image → Text** retrieval

### Metrics

We report standard retrieval metrics:
- Recall@1 (R@1)
- Recall@10 (R@10)
- Mean Reciprocal Rank (MRR)
- Median Rank (MedR)

## How to Use

```python
from datasets import load_dataset

# Load the dataset
dataset = load_dataset("SUSTech/Mars-VL-Pairs")

# Access a sample
print(dataset["train"][0]["refined_caption"])
```

For detailed instructions on the retrieval-centric protocol and official evaluation scripts, please refer to our [Official Dataset Documentation](https://github.com/ml-stat-Sustech/MarsRetrieval/blob/main/docs/DATASET.md).

## Citation

If you find this useful in your research, please consider citing:

```bibtex
@article{wang2026marsretrieval,
title={MarsRetrieval: Benchmarking Vision-Language Models for Planetary-Scale Geospatial Retrieval on Mars},
author={Wang, Shuoyuan and Wang, Yiran and Wei, Hongxin},
journal={arXiv preprint arXiv:2602.13961},
year={2026}
}
```