Dataset Viewer
Auto-converted to Parquet
text
stringlengths
94
444
images
sequencelengths
1
1
"Comparing panels a, b, c, and d, which statement best describes the data variance? <|vision_start|>(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What is the duration of the course mentioned in the image? <|vision_start|><|image_pad|><|vision_en(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What is the primary purpose of the PTC in lithium batteries? <|vision_start|><|image_pad|><|vision_(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What is the date?\nYour answer should be very brief. <|vision_start|><|image_pad|><|vision_end|>\nR(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What is Bert Shulimson's title?\nYour response must be concise. <|vision_start|><|image_pad|><|visi(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"How is the baseline CO2 emissions calculated for affected EGUs in the low load natural gas-fired or(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What are some suggestions Liberty Medical Group should consider to improve their accounts receivabl(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What does the sequence of transformations \\( U_{D_\\sigma,11}, U_{D_\\sigma,9}, U_{D_\\sigma,7}, U(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"What measures will the Secretary determine to assess the quality of care furnished by the ACO? <|vi(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
"How is the CT kerma index measured? <|vision_start|><|image_pad|><|vision_end|>\nRepresent the give(...TRUNCATED)
["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLD(...TRUNCATED)
End of preview. Expand in Data Studio

Copali train split used in MoCa Continual Pre-training

🏠 Homepage | πŸ’» Code | πŸ€– MoCa-Qwen25VL-7B | πŸ€– MoCa-Qwen25VL-3B | πŸ“š Datasets | πŸ“„ Paper

Introduction

This is a interleaved multimodal pre-training dataset used in the modality-aware continual pre-training of MoCa models. It is adapted from Copali and its corpus by concatenating queries and positive documents.

The dataset consists of interleaved multimodal examples. text is a string containing text while images are image binaries that can be loaded with the following code snippet:

import PIL.Image
from io import BytesIO

image_bytes = example['images'][0]
image = PIL.Image.open(BytesIO(image_bytes))

Citation

MoCa

@article{chen2025moca,
  title={MoCa: Modality-aware Continual Pre-training Makes Better Bidirectional Multimodal Embeddings},
  author={Chen, Haonan and Liu, Hong and Luo, Yuping and Wang, Liang and Yang, Nan and Wei, Furu and Dou, Zhicheng},
  journal={arXiv preprint arXiv:2506.23115},
  year={2025}
}

Colpali

@inproceedings{faysse2024colpali,
  title={Colpali: Efficient document retrieval with vision language models},
  author={Faysse, Manuel and Sibille, Hugues and Wu, Tony and Omrani, Bilel and Viaud, Gautier and Hudelot, C{\'e}line and Colombo, Pierre},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2024}
}
Downloads last month
149