datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
pharma-translational-biomarker-validity-v0.1What this repo is for
decide if a biomarker is worth using as a decision driver
stop teams using biomarkers that only correlate
prevent false confidence in early trials
choose enrichment markers for patient selection
choose activity markers for dose finding
reduce wasted phase 2 and phase 3 spend
What the model must do
You give it one row.
It must output one label.
coherent
incoherent
How to interpret the label
coherent means
biomarker sits on the causal chain
biomarker predicts outcome… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/pharma-translational-biomarker-validity-v0.1.clinical-healing-trajectory-permissible-transition-rhythm-validity-v0.1What this dataset tests
Whether a model can judge if a healing phase sequenceobeys permissible transition rules and rhythm patterns.
Required outputs
transition_validity
rhythm_class
off_grammar_flag
Transition validity labels
permissible
questionable
invalid
Rhythm labels
biphasic
oscillatory
plateau_then_jump
steady_linear
decline_then_rebound
chaotic_drift
Off-grammar flag
off_grammar_yes
off_grammar_no
Permissible transition assumptions
acute_drop… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-healing-trajectory-permissible-transition-rhythm-validity-v0.1.cig-intervention-validity-mapping-v0.1What this dataset tests
Whether an intervention produces only licensed effects.
No extra effects.No missing effects.No reverse causation.
Why this exists
Models often narrate causality.
This benchmark enforcesintervention discipline.
Labels
valid-intervention
overreach
underreach
confounded
Typical failures
non-descendant changes
missing downstream effects
parent altered by child intervention
collider misuse
Suggested prompt wrapper
System
You evaluate whether an… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/cig-intervention-validity-mapping-v0.1.cross-domain-transfer-validity-stress-test-v0.1What this dataset tests
Whether a model can stress-test a cross-domain transfer claimby identifying invalidity risks, confounders, and a minimal validation plan.
Required outputs
invalid_transfer_risks
confounder_list
minimal_validation_plan
transfer_confidence_0_100
Typical failures
no boundary conditions
confounders not named
validation plan too broad to run
confidence score without justification
Suggested prompt wrapper
System
You stress-test a cross-domain… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/cross-domain-transfer-validity-stress-test-v0.1.face_validity_synth
Dataset Card
Dataset Description
This is a synthetic dataset that I am using to try to figure out how to make a widget. Watch this space.
from datasets import load_dataset
ds = load_dataset("emmharv/face_validity_synth", split="train")
ds.shuffle(seed=42).select(range(5))