Discrete MeanFlow: Efficient One-Step Generation Model
Discover Discrete MeanFlow, a novel one-step generative model using conditional transition kernels for efficient discrete state space generation.
Understanding Emergent Misalignment in LLM Fine-Tuning
Explore how data-mediated transfer causes emergent misalignment in large language models during fine-tuning on harmful datasets.
Adaptive Smooth Tchebycheff for Multi-Objective Policy Optimization
Discover the Adaptive Smooth Tchebycheff method improving multi-objective policy optimization in reinforcement learning for robotics and complex tasks.
WriteSAE: Advanced Sparse Autoencoders for Recurrent Models
Discover WriteSAE, a novel sparse autoencoder boosting recurrent model performance with efficient matrix cache write techniques.
Multi-Quantile Regression Boosts Extreme Rainfall Prediction
Discover how multi-quantile regression improves extreme precipitation downscaling, enhancing flood risk prediction with advanced machine learning models.