STARJOB: DATASET FOR LLM-DRIVEN JOB SHOP SCHEDULING
Keywords: JSSP, Large Language Models, supervised dataset, Starjob, artificial intelligence, sampling method, LLM
TL;DR: We introduce the very first supervised dataset specifically designed to train LLMs for JSSP. Surprisingly, our findings demonstrate that LLM-based scheduling can achieve performance comparable to other neural approaches.
Abstract: The Job Shop Scheduling Problem (JSSP) presents a significant challenge in opti-
mizing production processes. This problem requires efficient allocation of jobs to
a limited number of machines while minimizing total processing time (makespan).
Although recent advancements in artificial intelligence have produced promising
solutions, such as reinforcement learning and graph neural networks, this paper
investigates the potential of Large Language Models (LLMs) for addressing JSSP.
We introduce the first supervised 120k dataset called Starjob specifically designed
to train LLMs for JSSP and we subsequently fintune the LLaMA 8B model on
this dataset using Lora. We compare the average makespan gap of our end-to-
end LLM-based scheduling method with that of the most widely used priority
dispatching rules (PDRs) and neural methods such as L2D. Surprisingly, our find-
ings indicate that LLM-based scheduling not only surpasses traditional PDRs but
also achieves on average 11.28% on DMU and 3.29% gap improvement on the
Tailard benchmarks compared to the state-of-the-art L2D method.
Supplementary Material: pdf
Primary Area: datasets and benchmarks
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 11565
Loading