Original title: Cooperative Multi-Task Semantic Communication for Joint Classification and Regression Tasks
Ahmad Halimi Razlighi, Mohammad Siddiqur Rahman, Maximilian H. V. Tillmann, Edgar Beck +1 · September 2026
Future smart networks need to send meaning rather than perfect copies of raw data. Doing this for several different jobs at once is tricky. Earlier setups only tested matching, simple jobs on basic data. In this paper, researchers improved their cooperative communication system to tackle mixed jobs on a complex real-world dataset called Cityscapes. These jobs include sorting items into categories and predicting continuous numbers. The system splits encoding work between one shared unit and several specialized units. It uses a rule called information maximization to handle both discrete and continuous information together. Tests showed that this cooperative design clearly beats standard digital sending, doing each task alone, and using one encoder for multiple decoders.
Imagine organizing a small kitchen where you need to cook and bake at the same time. The hard part is that utensils and ingredients overlap, but each task has its own steps, so using everything for one task leaves the other unprepared. In this paper, the system has a shared shelf for common tools like knives and mixing bowls, and separate drawers for task-specific items like baking pans or frying pans. By keeping shared resources available but specialized ones handy, it handles both cooking styles at once without waste.
Future intelligent networks like self-driving cars need to process multiple types of information at once. This work makes that more efficient, so devices can understand scenes better without sending all the data.
The abstract does not explicitly state limitations, but it indicates that the framework is evaluated only on the Cityscapes dataset, so its performance in other real-world scenarios is not confirmed. It also builds on prior work, so the novelty might be incremental rather than foundational.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Cooperative Multi-Task Semantic Communication for Joint Classification and Regression Tasks — Ahmad Halimi Razlighi, Mohammad Siddiqur Rahman, Maximilian H. V. Tillmann et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.