Huawei · 9.1.3

Huawei argues that generalized performance in CSI prediction can be achieved with mixed datasets (Generalization Case 3), thereby proposing that no explicit NW indications or associated IDs are needed for training/inference consistency. They propose reusing Rel-18 MIMO mechanisms for CSI report configuration and suggest investigating new priority rules where monitoring reports take precedence over inference reports. For performance monitoring, Huawei prefers SGCS as the primary metric and advocates for L1 signaling for monitoring results due to latency constraints, while proposing separate CSI report configurations for inference and monitoring processes. Regarding LCM, they propose separating CPU counting between legacy and AI/ML CSI reporting, introducing activation delays into processing timelines based on functionality state, and implementing memory occupancy alignment mechanisms. Finally, they support reusing the Beam Management functionality alignment signaling procedure (Steps 1-5) for CSI prediction, including reporting UE requirements for CPU, memory, and timeline in Step 4.

Huawei presents a technical case against the necessity and feasibility of introducing associated IDs or other NW-side indications for CSI prediction consistency. They argue that the massive number of impacting factors, such as antenna layout and down tilt angles, makes categorization by the network difficult and risks proprietary disclosure of network planning information. Huawei contends that unlike beam management or positioning, CSI prediction models achieve sufficient generalized performance using mixed datasets (Generalization Case 3), rendering explicit consistency indications unnecessary. Instead, they propose that consistency be ensured through UE-side performance monitoring in an implementation manner, allowing the UE to select the best-matching model without network signaling overhead.

Huawei advocates AGAINST introducing network-side associated IDs or explicit indications for CSI prediction consistency, arguing that such mechanisms are both infeasible (due to massive impacting factors and network burden) and unnecessary (since generalized models work well with mixed datasets). They push FOR UE-side performance monitoring as an implementation-based solution, distinguishing CSI prediction from beam management and positioning use cases where associated IDs were deemed necessary.