AKS Service Load Balancing Strategy

[smvdn]

What is the load balancing strategy is implemented for the Service to the backend Pods in AKS kube-proxy ? Is it random or least connection or round robin ?
Recommend to include this in AKS Docs. At least I am not able to find this from AKS documentation

PS: Original Kubernetes Doc for the same https://kubernetes.io/docs/concepts/services-networking/service/#virtual-ips-and-service-proxies