ekacare/MedMCQA-Indic

MedMCQA - Indic translations

The 4,183 labeled questions of MedMCQA (openlifescienceai/medmcqa, validation split - the official test split has no public answers) translated into 11 Indian languages, published as one test split per language config. Ids, answer keys (cop), subjects and topics are identical across all 12 configs, so scores are directly comparable across languages.

config	language	rows
en	English	4183
hi	Hindi	4183
bn	Bengali	4183
gu	Gujarati	4183
kn	Kannada	4183
ml	Malayalam	4183
mr	Marathi	4183
or	Odia	4183
pa	Punjabi	4183
ta	Tamil	4183
te	Telugu	4183
as	Assamese	4183

How the translations were made

• Translator: claude-opus-5-5 (Claude Opus 5.5). The question and its 4 options are translated.
• exp (explanation, present on 2,206 rows) is kept in the original English in every config.

Prompts (prompts/)

• prompts/mcq_translate_v3__0cc4b4682dfd.txt - the system prompt used for every translation (the 11-language form; its name is the prompt_template_id in the data). Each call's user message holds a few source items as <item id="N"><question>..</question><opa>..</opa>..<opd>..</opd></item>; the model returns one <item id="N" lang="xx"> block per item x language with the same fields.
• prompts/mcq_translate_v3_revision_note.txt - appended (to the single-language form of the same prompt) for the one integrity-fix re-translation (prompt_template_id ending +fix1); the user message then also carries the previous translation and the failed automatic checks.
• prompts/mcq_judge_v1__4c260e8a77f6.txt - the MQM judge used to tune the prompt ({LANG} = target language). Not run on every row.

Fields

MedMCQA columns (id, question, opa, opb, opc, opd, cop, choice_type, subject_name, topic_name, exp) + language, translator, prompt_template_id, integrity_flags (JSON; automatic checks that fired on that row).

License and attribution

Released under the Apache License 2.0, the same license as the source dataset.

This is a derivative of MedMCQA (openlifescienceai/medmcqa, Pal, Umapathi & Sankarasubbu, 2022), licensed under Apache-2.0. Changes made: the question, opa, opb, opc and opd fields of the 4,183 validation-split rows were machine-translated into 11 Indian languages and the language, translator, prompt_template_id and integrity_flags columns were added; all other fields (including exp) and the en config are the original data, unchanged.

Citation

If you use this dataset, please cite it:

@misc{ekacare2026medmcqaindic,
  title        = {MedMCQA-Indic: MedMCQA Translated into 11 Indian Languages},
  author       = {{Eka Care}},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/ekacare/MedMCQA-Indic}}
}

and the original MedMCQA dataset:

@InProceedings{pmlr-v174-pal22a,
  title     = {MedMCQA: A Large-scale Multi-Subject Multi-Choice Dataset for Medical domain Question Answering},
  author    = {Pal, Ankit and Umapathi, Logesh Kumar and Sankarasubbu, Malaikannan},
  booktitle = {Proceedings of the Conference on Health, Inference, and Learning},
  pages     = {248--260},
  year      = {2022},
  volume    = {174},
  series    = {Proceedings of Machine Learning Research},
  publisher = {PMLR},
  url       = {https://proceedings.mlr.press/v174/pal22a.html}
}