PubMedCausal

Key facts

Domain

Biomedical literature Modern medicine

Content

Annotated corpora · LiteratureSpan-level annotated corpus for causal relation extraction in biomedical text, built from 30,000 paragraph-level PubMed abstract samples with 6,491 adjudicated cause-effect pairs.

Size

—As stated: 30,000 paragraph-level samples

Language / modality

English / Text

Publisher

jaypee01

Year

—

Expert review

—

Verification

Verified: license text seen

Commercial use of PubMedCausal

Commercial use: Yes. License status: Permissive.

Conditions

Notice kept

Quoted statements and their sources are in PubMedCausal license evidence.

Verdicts record what sources state and are not legal advice.

Tags

Task format

Relation extraction

Quoted from the source**PubMedCausal** is a span-level annotated corpus for **causal relation extraction in biomedical text**.Snapshot s01

Text classification

Quoted from the source- text-classificationSnapshot s01

Intended use

Evaluation

Quoted from the sourceThese properties make the corpus suitable for evaluating causal extraction under conditions that more closely resemble causal statements encountered in biomedical literature.Snapshot s01

Fine-tuning

Specialty

Multi-specialty

Source type

Research literature

Quoted from the sourceThe corpus is derived from PubMed abstracts.Snapshot s01

Expert involvement

Reviewed, method unspecified

Quoted from the sourceAnnotations were subsequently reviewed and adjudicated to produce the final corpus.Snapshot s01

Official split

Train and test splits

Quoted from the sourceThe Hugging Face release provides train and test partitions for the supported task configurations.Snapshot s01

Author-statedInferredHuman-confirmed

PubMedCausal license evidence 1

1. Hugging Face card metadataPage opened
   MIT

   Third-party text · Hugging Face card metadatalicense: mit

   Statement found in snapshot s01 (readme)

   huggingface.co2026-11-12 · s01

Source and lineage

No upstream or downstream datasets recorded.

Known issues with PubMedCausal

• Severe class imbalance: only 13.15% of samples contain annotated causality, so accuracy alone is misleading.
• Annotated relations express causal claims in the literature; they do not establish that the claims are true, and predictions are not medical advice.

Downloading PubMedCausal and API access

No file on Mdexa (see the verdict). Get it from the source and confirm the terms there:

• huggingface.co/datasets/jaypee01/PubMedCausal

curl -L --create-dirs -o ".gitattributes" "https://huggingface.co/datasets/jaypee01/PubMedCausal/resolve/fb01de12ea3d12fe45a5a5ba173c8d113a00e949/.gitattributes"
curl -L --create-dirs -o "PubMedCausaldata/30k_test.json" "https://huggingface.co/datasets/jaypee01/PubMedCausal/resolve/fb01de12ea3d12fe45a5a5ba173c8d113a00e949/PubMedCausaldata/30k_test.json"
echo "ee4db1cea406a7387ba6696310402c5c2eb31a2cc584e42419117446c6f9e68b  PubMedCausaldata/30k_test.json" | sha256sum -c
curl -L --create-dirs -o "PubMedCausaldata/30k_train.json" "https://huggingface.co/datasets/jaypee01/PubMedCausal/resolve/fb01de12ea3d12fe45a5a5ba173c8d113a00e949/PubMedCausaldata/30k_train.json"
echo "4cd866724d3624099f73237f1af1ec959a5a2744c607f76fc9862606459ef9b7  PubMedCausaldata/30k_train.json" | sha256sum -c
curl -L --create-dirs -o "PubMedCausaldata/validation_set.json" "https://huggingface.co/datasets/jaypee01/PubMedCausal/resolve/fb01de12ea3d12fe45a5a5ba173c8d113a00e949/PubMedCausaldata/validation_set.json"
curl -L --create-dirs -o "README.md" "https://huggingface.co/datasets/jaypee01/PubMedCausal/resolve/fb01de12ea3d12fe45a5a5ba173c8d113a00e949/README.md"

Citing PubMedCausal

Mdexa, PubMedCausal, https://mdexa.com/datasets/pubmedcausal/, verified 2026-11-14.

Machine-readable

pubmedcausal.jsonpubmedcausal.mdcatalog.json

Listed in

• Data reviewed by clinicians or experts
• Datasets with official train and test splits
• Medical datasets for evaluation that have a permissive license
• Medical datasets for fine-tuning that have a permissive license

Datasets similar to PubMedCausal

• CBLUEVerified by MdexaNamed entity recognitionEvaluation UnstatedPending 8 tasks 2026-11-12

  License unstated: Only a code repository license; no data license stated; link only until resolved.

  Clinical medicine · Question banks and benchmarks · Annotated corpora

  Open the cardSave

  Shared: Task format Relation extraction, Text classification · Intended use Evaluation · Specialty Multi-specialty · Official split Train and test splits · system Modern medicine · content Annotated corpora

• PMC-Patients-ReCDSVerified by MdexaText matchingEvaluation RestrictedLink only — 2026-11-14

  License restricted: restrictions are read broadly: CC-BY-NC-SA-4.0. Risk flag(s): derived, Real patient text. Link only.

  Clinical medicine · Biomedical literature · Annotated corpora · Clinical cases · Literature · Derived · Real patient text

  Open the cardSave

  Shared: Intended use Evaluation · Specialty Multi-specialty · Source type Research literature · Official split Train and test splits · domain Biomedical literature · content Annotated corpora, Literature

• PromptCBLUEVerified by MdexaNamed entity recognitionEvaluation RestrictedLink only 68.9K items 2026-11-12

  License is restricted: README prohibits commercial use; link only.

  Clinical medicine · Question banks and benchmarks · Annotated corpora · Derived

  Open the cardSave

  Shared: Task format Relation extraction, Text classification · Intended use Evaluation · Specialty Multi-specialty · Official split Train and test splits · system Modern medicine · content Annotated corpora

• PubMedQAVerified by MdexaMultiple-choice (single answer)Evaluation PermissiveLink only — 2026-11-14

  License permissive: first-hand data-scope statement(s) allow commercial use and redistribution. Link only.

  Biomedical literature · Clinical medicine · QA, dialogue and instruction data · Annotated corpora

  Open the cardSave

  Shared: Intended use Evaluation · Specialty Multi-specialty · Source type Research literature · domain Biomedical literature · content Annotated corpora

• IMCS-21Verified by MdexaPediatricsMulti-turn dialogueEvaluation UnstatedPending 4,116 records 2026-11-12

  License unstated: No license on the repository page; link only until resolved.

  Clinical medicine · QA, dialogue and instruction data · Annotated corpora · Real patient text

  Open the cardSave

  Shared: Task format Text classification · Intended use Evaluation · system Modern medicine · content Annotated corpora