Dxbench: A Benchmark for Interpretable Medical Diagnosis Agents

The Dxbench dataset is associated with the 2024 arXiv paper 'CoD, Towards an Interpretable Medical Agent using Chain of Diagnosis'. It was uploaded to Hugging Face by FreedomIntelligence on 2024-08-23. The platform tags indicate it contains clinical text in English and Chinese, formatted in JSON, and is intended for token classification tasks.

Use Cases

• Training and benchmarking medical diagnosis agents based on the 'Chain of Diagnosis' concept mentioned in the description.
• Developing interpretable AI models for clinical text analysis based on the dataset's stated domain.
• Conducting token classification tasks on medical and healthcare text as indicated by the platform tags.

Strengths

• Dataset is directly linked to a 2024 peer-reviewed arXiv paper (2407.13301), providing academic context.
• Platform tags specify dual-language content (English and Chinese) and a structured JSON format.
• Last updated on 2024-08-23, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file size, and specific data volume are unknown, which may limit suitability assessment.
• The license is unspecified, requiring verification before use.

Provenance

Source

FreedomIntelligence, associated with the authors of the arXiv paper 'CoD, Towards an Interpretable Medical Agent using Chain of Diagnosis'.

Collection Method

Likely created as a benchmark for the research described in the associated paper.

Time Range

null

Freshness

Last updated 2024-08-23 04:56:45.

Geography

Platform tags indicate a US region, but specific geographic coverage of the data is unknown.