Starjob: Dataset for LLM-Driven Job Shop Scheduling

Henrik Abgaryan, Tristan Cazenave, Ararat Harutyunyan · 2025

Read Starjob: Dataset for LLM-Driven Job Shop Scheduling by Henrik Abgaryan, Tristan Cazenave, Ararat Harutyunyan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-01877.txt.

Description

Abgaryan, Cazenave, and Harutyunyan present a paper from twenty twenty-five on a dataset for language-model job shop scheduling. Scheduling jobs on machines to cut total time is a classic hard problem. Their dataset holds one hundred thirty thousand solved-style instances. A mid-size language model is fine-tuned on it cheaply. The tuned model schedules end to end from problem text. Standard benchmarks test the approach against classic rules. It beats rule-based dispatching across the board.

It also tops a leading neural scheduler on two benchmark suites. Average gains reach double digits on one suite. The authors claim large untapped potential for language models here. Combinatorial optimization gains a new test bed.

Catalog record

CREATORHenrik Abgaryan, Tristan Cazenave, Ararat Harutyunyan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-01877

Browse the full library · About the Tower of Babel