Adapt PRISM to Within-Run Semantic Association Drift

Develop online adaptation mechanisms for PRISM when semantic association statistics change during a learning run, including changes in user participation, transmission behavior, or semantic-generation patterns.

Background

The evaluation fixes each association matrix during an individual run and varies it only across separate runs. Consequently, PRISM is assessed under stationary association statistics during learning, even though semantic contributions can vary with segment recency and access outcomes. The authors explicitly identify adaptation to within-run drift as unresolved, which is important for deployment in dynamic 6G scenarios where users may join or leave or alter their transmission and semantic-generation patterns during operation.

References

Finally, each association matrix is fixed within a run and varied across runs; hence, the evaluation assumes stationary association statistics during learning. Instantaneous semantic contributions still vary with $\delta_kt$ and access outcomes. Adaptation to within-run drift is left for future work.

— A Semantic-Aware Multiple Access Scheme Leveraging Spatial Redundancy for Uplink-Dominant Network Services  (2609.03559 - Mazandarani et al., 3 Sep 2026) in Section 5, Evaluation, immediately after the simulation-configuration discussion