NVIDIA Research Projects
Official@nvlabs
Offers specialized frameworks for robot manipulation benchmarking, navigation training orchestration, and vision-language model fine-tuning.
Agent Skills by NVIDIA Research Projects
Showing 8 vetted skills indexed across 3 GitHub repositories.
NVlabsrobolab-taskgen
Generate Python robot manipulation task definitions for the RoboLab benchmark framework.
NVlabsrobolab-scenegen
Generate USD scene files from natural language descriptions for robot manipulation benchmarks.
NVlabsalpamayo1-5-sft
Automate Alpamayo-1.5 VLM fine-tuning on PAI or LingoQA datasets.
NVlabsalpamayo1-sft
Automate end-to-end fine-tuning of the Alpamayo-1 VLM on the PAI dataset.
NVlabsalpamayo1-x-rl
Automate RL post-training for Alpamayo VLM action models with Cosmos-RL and GRPO.
NVlabscompass-doctor
Diagnose COMPASS framework setup and training failures via automated environment validation checks.
NVlabscompass
Orchestrate COMPASS robot navigation training from SAGE-10k scene search to OSMO cloud submission.
NVlabscompass-newembodiment
Generate COMPASS configuration files and register new robot embodiments.
Frequently Asked Questions About NVIDIA Research Projects
FAQPage SchemaWhat specific research tasks are enabled by these frameworks?▼
These frameworks enable the generation of robot manipulation task definitions, the creation of USD scene files for simulation, the registration of new robot embodiments, and the end-to-end fine-tuning of vision-language models using PAI or LingoQA datasets.
Which personas benefit from using these research repositories?▼
These repositories are designed for robotics researchers, simulation engineers, and machine learning practitioners focused on embodied intelligence, navigation training, and vision-language model optimization within the NVIDIA research ecosystem.
What are the primary dependencies for running these training frameworks?▼
Execution requires access to the COMPASS framework for navigation training, the RoboLab benchmark suite for manipulation tasks, and specific infrastructure support for PAI datasets, Cosmos-RL, and GRPO-based reinforcement learning post-training.