NeurIPS Must Enforce AI Safety Reproducibility Standards
NeurIPS should adopt strict reproducibility standards to validate frontier AI safety claims, ensuring transparency and trust in advanced AI deployments.
Robust OOD Detection with Synergistic Score Smoothing
Discover ROSS, a robust out-of-distribution detection framework using median smoothing to enhance accuracy and defend against adversarial attacks.
Neuroscience Insights on Visual Interest in Multimodal AI
Explore how neuroscience principles shape visual interest representation in multimodal transformers, enhancing AI alignment with human cognition.
Entropy Minimization for Test-Time Adaptation in Autoregressive Models
Explore a unified entropy minimization framework enhancing test-time adaptation in autoregressive models, validated on Whisper ASR across 20+ domains.
Robotic Service Governance: Ensuring Admissible Reconfiguration
Explore how RoSO and SMGI frameworks enable admissible reconfiguration and robust governance in evolving robotic service systems.