datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
reasoning-chain-dependency-integrity-v0.1
What this dataset tests
Whether each step in a reasoning chain properly depends on prior steps.
The chain itself is the object.
Not the final answer.
Why this exists
Models often reach correct conclusions using:
skipped premises
reordered logic
circular justification
post-hoc explanations
This dataset isolates those failures.
Data format
Each row contains:
premises
reasoning_steps
dependency_graph
claimed_conclusion
The dependency graph… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/reasoning-chain-dependency-integrity-v0.1.clinical-reasoning-boundary-integrity-v0.1Clinical Reasoning Boundary Integrity v0.1
Purpose
Capture failures where models continue reasoning beyond what clinical evidence supports.
The dataset focuses on boundary discipline in clinical interpretation rather than generic accuracy.
It detects when models:
• extend conclusions past trial endpoints
• generalize beyond studied populations
• overstate safety or efficacy
• continue recommendations after uncertainty
• present speculation as evidence
This is not a correctness dataset.
It is a… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/clinical-reasoning-boundary-integrity-v0.1.