SiMing-Bench: Evaluating Procedural Correctness from Continuous Interactions in Clinical Skill Videos

TL;DR

SiMing-Bench is a new benchmark for evaluating how well multimodal models understand procedural correctness in clinical videos by tracking ongoing interactions and state updates.

Contribution

It introduces a novel benchmark and dataset for assessing models' ability to judge procedural correctness from continuous interactions in clinical videos.

Findings

Models show weak agreement with physician judgments.

Global assessments overestimate models' procedural understanding.

Performance bottleneck is modeling interaction-driven state updates.

Abstract

Current video benchmarks for multimodal large language models (MLLMs) focus on event recognition, temporal ordering, and long-context recall, but overlook a harder capability required for expert procedural judgment: tracking how ongoing interactions update the procedural state and thereby determine the correctness of later actions. We introduce SiMing-Bench, the first benchmark for evaluating this capability from full-length clinical skill videos. It targets rubric-grounded process-level judgment of whether interaction-driven state updates preserve procedural correctness across an entire workflow. SiMing-Bench is instantiated with SiMing-Score, a physician-annotated dataset of real clinical skill examination videos spanning cardiopulmonary resuscitation, automated external defibrillator operation, and bag-mask ventilation, each paired with a standardized step-wise rubric and dual-expert labels. Across diverse open- and closed-source MLLMs, we observe consistently weak agreement with physician judgments. Moreover, weak performance on rubric-defined intermediate steps persists even when overall procedure-level correlation appears acceptable, suggesting that coarse global assessment substantially overestimates current models' procedural judgment ability. Additional analyses with binary step judgment and step-aligned clips indicate that the bottleneck is not merely fine-grained scoring or temporal localization, but modeling how continuous interactions update procedural state over time.