R1-2410192
discussion
Other aspects of AI/ML model and data
Summary
Continental Automotive presents their position on AI/ML model identification and transfer mechanisms for NR air interface, proposing 4 specific enhancements. The document focuses on configurable mapping relations for model-related information exchange and efficient parameter transfer across different model structures.
Position
Continental advocates FOR configurable mapping relations as the primary mechanism for model identification and parameter transfer, strongly supporting the MI-Option2 framework with ID-X transmission from network to UE. They push FOR leveraging model structure similarity to enable partial parameter transfer and pre-configured mapping relations to optimize efficiency across different model structures in two-sided AI/ML deployments.
Key proposals
- Proposal 1 (Sec 2.1): Introduce a configurable mapping relation that allows both NW and UE to effectively exchange model-related information, aligning with the MI-Option2 framework
- Proposal 2 (Sec 2.1): Study the configuration of transmitting dataset IDs
- Proposal 3 (Sec 2.2): Consider the similarity of known model structures among models with or without the catalogued model properties when there are more than one model for model transfer/delivery
- Proposal 4 (Sec 2.2): Study pre-configuration of mapping relation for efficient target parameter transfer across different model structures