June 2025 performance summary focused on advancing latency-sensitive scheduling through the Latency Hiding Scheduler across TensorFlow and XLA. Delivered scheduling preferences to enable heuristic-driven prioritization, resulting in more efficient scheduling decisions, improved resource utilization, and lower latency for critical workloads. Across tensorflow/tensorflow and Intel-tensorflow/xla, implemented per-schedule preferences, introduced a mechanism to set and use preference values in the scheduling graph, and added a new API to schedule computations with specified preferences and statistics. No major bug fixes were recorded in this period; the work centered on feature enhancements and cross-team collaboration to raise performance and scalability of the latency-hiding scheduling workflow.