ClarusC64/clinical-therapeutic-pathway-trajectory-manifold-construction-v0.1

What this dataset tests

Whether a model can represent treatment pathways as trajectories through a high-dimensional patient state space.

It is a movie dataset.

Each example encodes

• —baseline state vector
• —time-indexed state vectors
• —interventions and timing
• —modality deltas
• —trajectory shape features

Manifold summary labels

• —smooth_improvement
• —plateau
• —oscillatory
• —deterioration
• —mixed_uncertain

Typical failures

• —treating the pathway as a single snapshot
• —ignoring modality lags
• —missing on-off coupling after stop or restart
• —collapsing tradeoffs into a single metric

Suggested prompt wrapper

System

You extract a trajectory representation from a treatment pathway.

User

Baseline State {baselinestatevector}

Timepoints {timepoints}

Interventions {interventions}

State Vector Series {statevectorseries}

Return

• —one manifoldsummarylabel
• —one sentence evidence trace

Citation

ClarusC64 dataset family