--- pretty_name: Med-R2 Bench language: - en license: other size_categories: - 10K Level 2 is not part of this benchmark release. The level numbering follows the original Med-R2 task design. ## ๐Ÿง  Evaluation tracks | Track | Medical reasoning focus | Representative tasks | | --- | --- | --- | | **Level 1 ยท Image quality** | Detect whether an image has been corrupted or altered | Noise recognition, artifact recognition | | **Level 3 ยท Anatomy** | Identify organs and reason about location or erased evidence | Organ recognition, organ location, organ eraser identification | | **Level 4 ยท Lesions** | Reason about body parts, lesions, boxes, and lesion size | Body-part recognition, lesion recognition, multi-box reasoning, lesion size | | **Level 5 ยท Clinical reports** | Generate a report under localized or image-level interference | Box report, noise interference, artifact interference | | **Joint QA ยท Multi-step reasoning** | Answer a sequence from modality to anatomy, abnormality, and diagnosis | Four-step sequential visual reasoning | ## โš”๏ธ Adversarial diversity Med-R2 is designed to test robustness under controlled visual changes. Depending on the task, examples include: - โœ… **Positive cases:** the visual evidence and the question are consistent. - โž– **Neutral / normal cases:** the image provides no abnormal evidence for the queried condition. - โŒ **Negative / adversarial cases:** the question, region, or visual evidence is intentionally mismatched. - ๐ŸŸฅ **Box interventions:** correct boxes, wrong boxes, or localized regions that redirect attention. - ๐Ÿงน **Erased-region cases:** relevant anatomy or evidence is removed or occluded. - ๐Ÿ”Š **Noise and artifact interventions:** image-level corruption is added while the underlying medical content is controlled. This diversity makes it possible to measure not only average accuracy, but also whether a model remains evidence-grounded across positive, normal, and adversarial conditions. ## ๐Ÿ“ Repository layout ```text Med-R2-Bench/ โ”œโ”€โ”€ train/ โ”‚ โ”œโ”€โ”€ visual_evidence_Level1.jsonl โ”‚ โ”œโ”€โ”€ visual_evidence_Level3.jsonl โ”‚ โ”œโ”€โ”€ visual_evidence_Level4.jsonl โ”‚ โ”œโ”€โ”€ visual_evidence_Level5.jsonl โ”‚ โ””โ”€โ”€ visual_evidence_joint_qa.jsonl โ”œโ”€โ”€ test/ โ”‚ โ””โ”€โ”€ same five JSONL files as train/ โ””โ”€โ”€ Med-R2-Bench/ โ”œโ”€โ”€ dataset/ # Referenced generated and processed images โ””โ”€โ”€ images/ # Referenced visual-evidence and joint-QA images ``` The JSONL files are the benchmark annotations. The image files are released separately in the nested `Med-R2-Bench/` directory so that the `image_paths` field remains reproducible after downloading the repository. ## ๐Ÿงพ Annotation format Each JSONL line is one example. The exact fields depend on the evaluation track, but common fields include: | Field | Description | | --- | --- | | `image_paths` | Repository-relative image path, such as `Med-R2-Bench/dataset/...png` | | `question` | The visual question or report-generation instruction | | `options` | Multiple-choice options when the task is closed-ended | | `answer` | Gold answer for the example | | `task` | Fine-grained task name, such as `noise_recognition` or `lesion_recognition` | | `type` | Data condition or adversarial bucket | | `modality` | Imaging modality, when available | Level 5 report examples use an open-ended `answer` field rather than multiple-choice options. Joint QA examples include the sequential prompts and gold answers for each reasoning step. ## ๐Ÿš€ Loading the annotations ```python from datasets import load_dataset dataset = load_dataset( "BiTaNan/Med-R2-Bench", data_files={ "train": "train/*.jsonl", "test": "test/*.jsonl", }, ) print(dataset) print(dataset["test"][0]) ``` To resolve an image locally after downloading the repository: ```python from pathlib import Path repo_root = Path("/path/to/Med-R2-Bench") image_file = repo_root / example["image_paths"] ``` The image paths are intentionally relative and contain no private server prefixes. The public package does not include model weights or prediction results. ## ๐Ÿ“Š Evaluation code and prompts The benchmark-construction code, metric scripts, comparison scripts, and prompt notes are maintained in the companion GitHub repository: ๐Ÿ‘‰ [https://github.com/BiTaNannnnn/Med-R2-Bench](https://github.com/BiTaNannnnn/Med-R2-Bench) The Hugging Face repository is the **data release**. The GitHub repository is the **code and documentation release**. ## ๐Ÿ” Release and usage notes - The released annotations use English prompts and answers. - Absolute server paths have been removed from the public annotations. - Model weights, private experiment outputs, API keys, and training logs are not included. - Please check the licenses and terms of the underlying medical image sources before redistributing the data or using it in a new collection. - This benchmark is intended for research evaluation and does not provide medical advice or a clinical diagnostic service. ## ๐Ÿ“š Citation ```bibtex @inproceedings{medr2-2026, title = {Med-R2: An Adversarial Benchmark for Evidence-Grounded Reasoning in Medical VLMs}, author = {Ma, Wen and collaborators}, booktitle = {Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing}, year = {2026} } ``` โญ If you use Med-R2 Bench, please cite the paper and link back to both the dataset and the [GitHub repository](https://github.com/BiTaNannnnn/Med-R2-Bench).