Roughly one in three papers accepted at the 43rd International Conference on Machine Learning (ICML)2026 leaned on NVIDIA’s hardware or open models, a quiet signal that open infrastructure, not proprietary secrecy, now underpins how AI science gets done.
Among the leading contributors at this year’s conference was NVIDIA, with 74 papers accepted and according to the company around 2,000 accepted papers cited NVIDIA GPUs, and 145 cited NVIDIA Nemotron, a family of open models, including open datasets, as the foundation for a new research.
The company mentioned that hundreds of additional papers were built on NVIDIA’s open model families including Cosmos, Isaac GR00T, BioNeMo, covering applications in physical AI, robotics, autonomous vehicles and biomedical research.
“Robot world models emerged as one of the conference’s fastest growing research areas with papers like DreamDojo pushing the boundary of how AI systems learn to reason about and act in physical environments,” the company said.
“DreamDojo, for example, learns how the physical world behaves from human video and builds on NVIDIA Cosmos open frontier models to predict how a robot would handle objects and operate in environments it was never trained on. It lets researchers evaluate policies, plan actions and teleoperate a virtual robot, accelerating development without the costs and risks of physical deployment,” it further explained.
ICML 2026 kicked off on Monday at COEX Convention and Exhibition Center, Seoul, South Korea and will run until July 11. The conference opened with an expo and tutorial day on July 6. The main conference is scheduled for July 7-9, followed by workshops on July 10-11. This year’s conference received a total of 23,918 paper submissions, more than double of ICML 2025. Out of these 6,352 papers were accepted, resulting in an overall acceptance rate of 26.6%.
Several papers were based on NVIDIA BioNeMo open models which helped researchers understand protein function, molecular behavior and genetic code.
“Papers like FLIP2 introduce public benchmarks for testing how well AI predicts the effects of protein mutations while KERMT is a new BioNeMo open model for predicting molecular properties important to drug discovery,” the company stated.
It also informed that synthetic data generation (SDG) drew significant attention this year at ICML with several Nemotron and physical AI open datasets, reflecting a broader shift in how researchers are thinking about training at scale without relying solely on human-labeled data.
Accepted papers show NVIDIA Nemotron being used not simply as a standalone model, but as a broader research stack comprising open model weights, datasets and training recipes for reasoning, tool use, safety, data curation and efficient inference.
Nemotron is a family of open artificial intelligence models, datasets, and training resources developed by NVIDIA to support research and enterprise AI development. Rather than just being a single model, it is an ecosystem designed to help researchers and developers build AI systems.
Also Read: ICML 2026 Opens in Seoul: Record 23,918 Submissions, New AI Review Rules
