R1-2409856
discussion
Discussion on other aspects of AI/ML model and data
From NEC
Summary
NEC's contribution discusses AI/ML model and data aspects for NR air interface, focusing on model identification procedures, model transfer methodologies, and UE capability reporting. The document contains 11 proposals and 10 observations covering model identification for one-sided models, inter-vendor collaboration for CSI compression, and model management procedures.
Position
NEC advocates FOR supporting model identification for one-sided models using associated IDs, Alternative B for model transfer methodology z4 (network-initiated signaling approach), RRC-based model parameter transfer with UE readiness indication, and UE failure reporting with cause values. They are AGAINST the monitoring-based consistency approach due to high delays and detailed UE internal condition reporting due to proprietary information disclosure concerns.
Key proposals
- Proposal 1 (Model Identification): Clarify the support of model identification for one-sided model use cases, at least when associated ID is needed to ensure the consistency between model training and model inference
- Proposal 2 (Model Identification): RAN1 to select the model identification procedure based on the outcome of the study to address inter-vendor collaboration issues for CSI compression use case
- Proposal 3 (Conditions): One or more associated ID(s) can be attached to one same model ID to reflect different NW side additional conditions
- Proposal 5 (Conditions): To ensure the consistency within a cell and across multiple cells, support UE to feedback whether associated ID is needed, at least for model inference
- Proposal 7 (Model Transfer): Support Alt. B for model transfer methodology z4 with 4-step procedure where UE reports capability, NW indicates candidates, UE reports supported structures, and NW transfers parameters
- Proposal 8 (Model Transfer): Support RRC signaling for transfer of AI/ML model parameters from gNB to UE, with discussion on storing parameters during idle/inactive state
- Proposal 9 (Model Transfer): Support model structure within the first indication to be identified/associated with a model structure id and the model parameters in the second indication to be identified/associated with a model id value
- Proposal 10 (Model Transfer): At least to support offline engineering of the model at the UE side, support UE indication to network if and when the transferred AI/ML model is ready to run at UE
- Proposal 11 (UE Capability): Specify UE indication to network about its inability to run a configured/activated AI/ML model/functionality due to UE's internal condition along with a relevant cause value for the failure
- Proposal 4 (Conditions): For inference for UE-side models, to ensure consistency between training and inference regarding NW-side additional conditions, prioritize model identification alignment, model training at NW and transfer to UE, and information provision on NW-side additional conditions
- Proposal 6 (Conditions): Study the grouping of cells that can ensure the consistency within a subset of cells