M2KR Challenge: A Multimodal Retrieval Dataset for Image-to-Document Matching

M2KR-Challenge is a multimodal retrieval dataset created by Jingbiao and last updated on February 4, 2025. It contains 6.42k query samples with images and optional text, and a collection of 47.3k textual passages with associated web screenshots. The dataset is designed for image-to-document and image+text-to-document matching tasks.

Use Cases

• Benchmarking image-to-document retrieval models based on the 6.42k image queries and 47.3k text passages.
• Training image+text-to-document matching systems based on the optional text questions paired with images.
• Evaluating multimodal search algorithms using the provided web screenshot paths associated with textual passages.

Strengths

• Contains 6.42k query samples with images and optional text.
• Includes a substantial collection of 47.3k textual passages with associated web screenshots.
• Specifically designed for two defined multimodal retrieval tasks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the main subsets is known, but total dataset size and file formats are unknown.
• Data may reflect bias inherent to the source web screenshots and passages.

Provenance

Source

huggingface, author Jingbiao

Collection Method

Likely web-scraped or compiled for a retrieval challenge, given the mention of web screenshots.

Time Range

null

Freshness

Last updated 2025-02-04 21:23:36.

Geography

Region 'us' is indicated in platform tags, but specific coverage is unknown.