ClarusC64/clinical-multidoctor-coherence-convergence-detection-v0.1

What this dataset tests

Whether a model can locate the turn where collective reasoning converges on a coherent diagnostic narrative.

Required outputs

• —convergence_point
• —coherencescore0_100
• —dominant_narrative

Scoring anchors

• —narrative explains all key findings
• —minimal extra assumptions
• —objections resolved or absorbed
• —cross-specialty conflict closed

Typical failures

• —picking final diagnosis without locating the turn
• —scoring coherence without referencing objections
• —ignoring integrative narratives that absorb contradictions

Suggested prompt wrapper

System

You detect diagnostic coherence convergence.

User

Dialogue excerpt {dialogue_excerpt}

Key findings {keyfindingslist}

Objections {objection_list}

Return

• —convergence point as T#
• —coherence score 0-100
• —dominant narrative in one sentence
• —one sentence evidence

Citation

ClarusC64 dataset family