Papers of Note

When Does Multimodal Learning Help in Healthcare? A Benchmark on EHR and Chest X-Ray Fusion

Kejing Yin, Haizhou Xu, Wenfang Yao, Chen Liu, Zijie Chen, Yui Haang Cheung, William K. Cheung, Jing Qin

Introduces CareBench, a benchmark evaluating multimodal fusion of electronic health record data and chest radiographs across 14 fusion methods, 3 clinical tasks, and realistic missing-data conditions. The authors report that fusion confers the greatest benefit for modality-distributed diseases, that modality balancing contributes more than architectural complexity, and that multimodal models do not inherently improve algorithmic fairness.

mHC: Manifold-Constrained Hyper-Connections

Zhenda Xie, Yixuan Wei, Huanqi Cao, et al.

A framework addressing training instability and memory overhead in Hyper-Connections architectures by projecting the residual connection space onto constrained manifolds. The authors report improved scalability for large-scale model training.

Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning

Haozhe Wang, Qixin Xu, Che Liu, Junhong Wu, Fangzhen Lin, Wenhu Chen

Reports that reinforcement learning improves LLM reasoning through a two-phase hierarchy comprising initial low-level skill acquisition followed by high-level strategic planning. Introduces HICRA (Hierarchy-Aware Credit Assignment), which allocates credit to planning tokens.

Foundation Models for Medical Imaging: A Critical Review

An analysis of recent foundation model papers, examining the reported claims regarding generalization, data efficiency, and clinical applicability, and the evidence supporting each.

Algorithmic Bias in Healthcare: Key Papers and Lessons

A review of principal papers on bias in medical AI, from the cost prediction study reported by Obermeyer and colleagues to documented disparities in pulse oximetry.

Real-World AI Deployment: What the Literature Reveals

An analysis of published reports of clinical AI deployments, both successful and unsuccessful, and the implementation factors associated with each outcome.

External Validation in Medical AI: A Systematic Review

An analysis of why reported performance frequently does not reproduce on external validation, covering validation methodology and commonly observed sources of optimistic bias.