MedErtas recorded demonstration sources
Prepared 1 November 2026. These example selections are not clinical products.

MedCalc-Bench Verified (with tools)
Dataset: https://huggingface.co/datasets/nsk7153/MedCalc-Bench-Verified
Original benchmark: https://github.com/ncbi-nlp/MedCalc-Bench
Credit: MedCalc-Bench authors, NCBI/NLM BioNLP, and MedCalc-Bench-Verified contributors.
Licence: Creative Commons Attribution-ShareAlike 4.0 International.
https://creativecommons.org/licenses/by-sa/4.0/
Selected benchmark cases, prompts and reference answers are reproduced from
the evaluation inputs. The model completions are separate recorded MedErtas
outputs. Ertas selected cases and added interface labels and review notes;
the recorded input and completion text has not been rewritten. Redistributed
benchmark material and any adaptations of it remain under CC-BY-SA-4.0.
This licence applies to the benchmark material, not the entire Ertas website.

LongHealth
Source: https://github.com/kbressem/LongHealth
Copyright 2024 Keno Bressem and contributors. Apache License 2.0.
The full licence is supplied at /static/medertas/licenses/LongHealth-Apache-2.0.txt
The fictional records, questions and reference answers are selected from the
task1 evaluation inputs. The original prompt includes task instructions and
the question. Ertas added the example selection, labels and AI-review notes.
The complete original prompt is loaded on request; completions are unchanged.

MedHallu
Source: https://huggingface.co/datasets/UTAustin-AIHealth/MedHallu
Project: https://github.com/MedHallu/MedHallu
Credit: Shrey Pandit, Jiawei Xu, Junyuan Hong, Zhangyang Wang, Tianlong Chen,
Kaidi Xu, Ying Ding and contributors. Dataset card declares the MIT licence.
The full upstream licence is supplied at /static/medertas/licenses/MedHallu-MIT.txt
The dataset derives from PubMedQA. Original source attribution and references
remain with the upstream dataset. The selected prompts, candidate answers and
reference verdicts are benchmark content. Recorded MedErtas completions are
shown unchanged. Ertas added selections, interface labels and AI-review notes.

All demonstration outputs require independent verification and clinical review.
These notices do not imply endorsement by benchmark authors or institutions.