NVIDIA expands open AI models, datasets, and tools at NeurIPS, showcasing the Alpamayo-R1 VLA model for autonomous driving and over 70 innovative projects in AI reasoning, medical research, and AV development.
NVIDIA DRIVE Alpamayo-R1 integrates reasoning AI with path planning for level 4 autonomy, giving AVs common sense to navigate complex scenarios. Researchers can customize the model for benchmarking or experimental AV applications, with reinforcement learning showing significant improvements in reasoning capabilities.
NVIDIA releases Cosmos Cookbook for physical AI developers, featuring LidarGen for AV simulation, NuRec Fixer for artifact correction, Cosmos Policy for robot policies, and ProtoMotions3 for training digital humans and humanoid robots with realistic scenes.
NVIDIA Nemotron additions include multi-speaker speech AI models, reasoning-based AI safety model, and tools for reinforcement learning and synthetic dataset generation. Ecosystem partners like CrowdStrike and Palantir are using Nemotron and NeMo tools for agentic AI development.
NVIDIA researchers present language AI innovations at NeurIPS, with dozens of research papers advancing language models. Attendees can explore these innovations at the Nemotron Summit, featuring an opening address by Bryan Catanzaro, VP of applied deep learning research at NVIDIA.
Read more at NVIDIA: NVIDIA Advances Open Model Development for Digital and Physical AI
