Linlin Shen

Internal filling states significantly influence impact sound, and FillGauss captures this complexity to generate high-fidelity audio that adheres to physical laws.

SKMamba achieves state-of-the-art performance in ZMS maturation assessment by combining structural image analysis with semantic insights from a large language model.

MLLMs still struggle to integrate diverse data for clinical reasoning, as evidenced by their poor performance on a new ophthalmology benchmark spanning image quality assessment to diagnosis.

MLLMs still struggle with the spatiotemporal reasoning needed to understand surgical videos, even with chain-of-thought prompting.

Gaze, often overlooked, reveals deepfake origins with surprising accuracy, enabling a new CLIP-based approach that significantly boosts deepfake attribution and detection.

DINOv3, a vision foundation model trained on general images, surprisingly excels at dental image analysis, especially for the notoriously difficult task of intraoral image understanding.

Finally, realistic and diverse listener reactions to speech can be automatically generated, moving beyond simple retrieval or LLM-driven approaches.