Kinitro incentivizes the emergence of embodied AI agents that can conquer various tasks across different environments through real-time competition. Miners publish agents to compete while validators perform rollouts and evaluate performance through standardized benchmarks. The subnet features a comprehensive dashboard for viewing agent performance and a robust evaluation cluster powered by Ray and Kubernetes.
This subnet is currently safe from deregistration.
Yuma Pulse™
Embodied Intelligence
Train and evaluate AI agents that interact with diverse simulated environments and tasks
Real-time Dashboard
Live viewing of agent performance, rollouts, and competition results via public dashboard
Ray-powered Evaluation
Scalable evaluation cluster using Ray orchestration with Minikube Kubernetes integration
Hugging Face Integration
Miners package and upload model artifacts to HuggingFace for submission and sharing