SFT-then-RL Beats Mixed-Policy Methods in LLM Reasoning
Discover how the SFT-then-RL pipeline outperforms mixed-policy methods in large language model reasoning after fixing key optimization bugs.
Age-Specific Models Improve Hypoglycemia Classification in T1D
Explore how age-specific models enhance hypoglycemia classification and management in type 1 diabetes using advanced data analytics and CGM.
ESIA Framework for Accurate Pedestrian Intention Prediction
Discover ESIA, an energy-based spatiotemporal model improving pedestrian intention prediction for safer autonomous driving and enhanced interaction awarene...
Efficient Far-Field Anomaly Detection in Expressway Videos
Discover VIBES, a novel AI framework using Bayesian inference and VLMs for accurate, real-time far-field anomaly detection in expressway surveillance.
AIPsy-Affect: Keyword-Free Emotion Test for Language Models
Discover AIPsy-Affect, a 480-item keyword-free battery to accurately interpret emotion in language models without bias from emotional keywords.