The reference tooling is under testing.
What it does
The repository README describes the purpose as to "establish an evaluation framework and use it for the evaluation of scenarios collected for the FS_AI4Media study", with "common testbed architectures and anchors, metrics (e.g. AI/ML task metrics, feasibility/performance metrics), and specific details (such as test configuration and constraints) for each scenario evaluation". 3GPP TR 26.847 V19.0.0 clause 4.1 states: "A common set of architectures is assumed for the evaluation framework, unless specified otherwise, irrespective of the scenario." Its reference [16] is this repository ("5G-MAG repository for AI4Media evaluations").
Contributors & How to Contribute
What's Built
The mapping of boxes to the repository is for its maintainers to confirm; the split inferencing testbed of figure 4.2-1 is not drawn. Click the figure to enlarge it.
Repositories & Releases
Early Access: available to members first
Standards implemented
3GPPThe specifications below are listed as background context, not as a prerequisite reading list.
3GPP TR 26.927 (Study on Artificial Intelligence and Machine Learning in 5G media services) and its companion evaluation report TR 26.847, both Release 19, plus TR 22.874 (Study on traffic characteristics and performance requirements for AI/ML model transfer in 5GS).
See AI/ML Evaluation Framework: Specification list for version history and 5G-MAG's own standards contributions.
Developer Updates
Last Synced: 2026-10-27 02:34 UTC
Early Access: available to members first