Constructing and Evaluating Clinical Reasoning Trajectories for Medical Agent.

arXiv:2609.05090v2 Announce Type: replace Abstract: Evaluation of medical artificial intelligence agents remains predominantly answer-centric, assessing only the correctness of final outputs while overlooking the quality of intermediate reasoning. In clinical