R1-2407697
discussion
Discussion on other aspects of AI/ML model and data
From Spreadtrum
Summary
Spreadtrum presents their views on AI/ML for NR air interface general aspects including data collection, model transfer/delivery, and model identification for two-sided models. The document contains 4 proposals and 4 observations addressing deprioritization of certain model transfer cases and procedures for model identification.
Position
Spreadtrum advocates FOR deprioritizing model transfer cases z1 and z2 in Rel-19 due to cross-vendor collaboration burdens and proprietary design disclosure risks. They support mechanism 1a for UE data collection to avoid privacy exposure to network, and promote network-controlled model identification where NW assigns model IDs. They are AGAINST redundant procedures in MI-Option 1 and prefer focusing RAN1 scope on model content rather than signaling formats.
Key proposals
- Proposal 1 (Data collection): For data collection for UE-side model training, no discussion is needed in RAN1
- Proposal 2 (Model transfer): From RAN1 perspective, the model transfer/delivery Case z1 is deprioritized in Rel-19
- Proposal 3 (Model transfer): From RAN1 perspective, the model transfer/delivery Case z2 is deprioritized in Rel-19 for two-sided model
- Proposal 4 (Model identification): From RAN1 perspective, the following procedure is an example of MI-Option 1 for two-sided model with 5 steps from data collection signaling to model delivery with NW-assigned model ID
- Observation 1 (Model transfer): Whether to support Case Z4 depends on the progress of 9.1.4.1 multi-vendor issue achieved
- Observation 2 (Model structure): RAN1 should focus on the content of model structure while model format and/or signaling part is up to other WG
- Observation 3 (Model identification): There would exist redundant procedures for MI-Option1 to do model identification for two-sided use cases
- Observation 4 (Model identification): MI-Option 2/3/4 can be considered for two-sided use cases