A Long Su

Arash Pakbin, Aaron Su, Donald K. K. Lee et al.

Objective: realtime monitoring of invasive ventilation (iV) in intensive care units (ICUs) plays a crucial role in ensuring prompt interventions and better patient outcomes. However, conventional methods often overlook valuable insights embedded within clinical notes, relying solely on tabular data. In this study, we propose an innovative approach to enhance iV risk monitoring by incorporating clinical notes into the monitoring pipeline through using language models for text summarization. Results: We achieve superior performance in all metrics reported by the state-of-the-art in iV risk monitoring, namely: an AUROC of 0.86, an AUC-PR of 0.35, and an AUCt of up to 0.86. We also demonstrate that our methodology allows for more lead time in flagging iV for certain time buckets. Conclusion: Our study underscores the potential of integrating clinical notes and language models into realtime iV risk monitoring, paving the way for improved patient care and informed clinical decision-making in ICU settings.

Mengqi Yuan, Zilong Zhou, Xinzhuang Xiong et al.

Existing computer-use benchmarks fail to capture the realism, complexity, and long-horizon demands of real-world computer use, limiting their ability to reveal the limitations of frontier agents. We introduce OSWorld 2.0, a benchmark of 108 long-horizon computer-use workflows across everyday and professional tasks, designed to capture complex and challenging real-world phenomena. Each task represents a realistic end-to-end workflow that takes human users a median of about 1.6 hours to complete and requires an average of 318 tool calls with Claude Opus 4.7 using maximum thinking, compared with about 30 in OSWorld 1.0. OSWorld 2.0 targets challenge phenomena that are common in real workflows yet underrepresented in prior benchmarks, spanning interaction-design challenges such as streaming interaction and dynamic environments, as well as agent-pattern challenges such as cross-source reasoning, implicit-state inference, and visual-spatial precision. Tasks are grounded in authentic input artifacts and cross-referenced against realistic stateful user profile data, and include separate safety reports auditing safety-sensitive execution. Under our primary binary-completion metric at 500 steps, Claude Opus 4.8 with maximum thinking and batched tool calls scores best but still completes only 20.6% of tasks at a 54.8% partial score; GPT-5.5 is far more token-efficient yet plateaus near 13%. These results show that current agents are still far from professional-level computer use: rather than stumbling on basic GUI control or coding, they lose track of constraints, miss information that arrives mid-task, guess rather than ask the user, and skip verification, struggling most when a task hinges on hidden state they must recover.