R1-2410261
discussion
Discussion on other aspects of AI/ML model and data
From ETRI
Summary
ETRI presents their technical views on AI/ML model identification and data management for NR air interface, making 4 specific proposals and 1 observation focused on establishing many-to-many relationships between IDs and models, and incorporating inference time information into model structures.
Position
ETRI advocates FOR establishing many-to-many relationships between various identifiers (Associated ID, ID-X) and Model IDs across both MI-Option1 and MI-Option2, enabling flexible model training scenarios where single datasets can train multiple models and single models can operate across multiple environments. They strongly push FOR including inference time information in known model structures, arguing that computational complexity metrics alone are insufficient since inference time varies non-linearly with model structure and hardware configurations.
Key proposals
- Proposal 1 (Sec 2.2.1): Associated IDs can be configured and managed for each functionality
- Proposal 2 (Sec 2.2.1): For MI-Option1 for the two-sided model, Associated ID and Model ID can have a many-to-many relationship
- Proposal 3 (Sec 2.2.2): For MI-Option2 for the two-sided model, ID-X and Model ID can have a many-to-many relationship
- Proposal 4 (Sec 2.3.1): The 'known model structure(s)' in the model transfer/delivery Case z4 can include inference time information