Original title: Automated Priority Rule Design for the Resource-Constrained Project Scheduling Problem: A Large Language Model-Guided Population-Based Search
Jingyu Luo, Mario Vanhoucke, José Coelho · September 2026
Planning a project is hard when tasks must follow a set order and share limited tools or workers. The goal is to finish everything in the shortest possible time. People often use simple priority rules to choose which task goes next. These rules are fast and clear, but no single traditional rule works well for every project. To solve this, researchers used a large language model to design new rules automatically. In this system, the model writes a group of candidate rules. A computer tests them on sample training projects to see how quickly they finish the work. The model then uses this score feedback to rewrite and improve the rules. Once finished, the best rule runs instantly on new projects without extra help from the model. The resulting rules beat traditional rules and outperformed rules evolved by genetic programming using a similar search effort.
Imagine planning a family meal where several dishes must share the oven and counter space. It's tricky to decide which dish to prepare first to avoid delays, especially when everyone has different preferences. In this paper, the project scheduling is like coordinating kitchen tasks, and the LLM acts as a smart organizer that learns from past meals to create better rules for the order, so everything finishes quicker.
Companies scheduling projects can complete work faster and save costs. This matters because better rules improve efficiency in industries like construction or software development where deadlines and resources are tight.
The abstract does not specify if the rules generalize to all project sizes or types beyond those tested. It relies on training projects that might not cover all scenarios, and performance could vary with different search efforts or ablation changes.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Automated Priority Rule Design for the Resource-Constrained Project Scheduling Problem: A Large Language Model-Guided Population-Based Search — Jingyu Luo, Mario Vanhoucke, José Coelho
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.