RareDis-Sub: Rare Disease Medical Reasoning Evaluation Subset

RareDis-Sub is a rare-disease-focused evaluation subset released with the ACL 2026 Findings paper, 'Eliciting Medical Reasoning with Knowledge-enhanced Data Synthesis: A Semi-Supervised Reinforcement Learning Approach'. The dataset is hosted on Hugging Face by author 'tdlhl' and was last updated on April 14, 2026. It is associated with a synthetic training dataset and a model for medical reasoning tasks.

Use Cases

• Benchmarking medical reasoning models based on the rare-disease-focused evaluation subset.
• Evaluating the performance of NLP models on specialized medical text described in the associated paper.
• Training or fine-tuning models for rare disease diagnosis or information synthesis using the linked synthetic dataset.

Strengths

• Dataset is explicitly designed for evaluating medical reasoning, a high-stakes application area.
• Focus on rare diseases addresses a specific and potentially underserved niche in medical AI.
• Has a clear academic provenance, being released with a peer-reviewed ACL 2026 paper.

Limitations

• Description metadata is limited; actual data quality, size, and structure require manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.

Provenance

Source

tdlhl on Hugging Face, associated with the ACL 2026 paper.

Collection Method

Likely created as a curated evaluation subset, potentially derived from or related to a larger synthetic dataset.

Freshness

Last updated 2026-04-14 02:58:32; freshness should be verified.