ImageCLEFmed 2013 case-based retrieval task relevance judgments expansion via an MLLM-as-a-Judge approach

This dataset contains additional relevance judgments (qrels) for the ImageCLEFmed 2013 case-based retrieval task. These judgments were generated using an MLLM-as-a-Judge approach, leveraging the Gemini 1.5 Pro model to simulate human assessment and expand the original qrels of the task.

The dataset includes 543,625 new relevance judgments for the ImageCLEFmed 2013 case-based retrieval task.

Each row represents a single judgment and follows the format: {topic_number} 0 {article_DOI} {relevance}Where:- {topic_number} refers to the query topic.- {article_DOI} identifies the document.- {relevance} indicates the degree of relevance (0 for non-relevant, 1 for relevant).