[2604.00039] Transformers for Program Termination

Computer Science > Programming Languages arXiv:2604.00039 (cs) [Submitted on 25 Mar 2026] Title:Transformers for Program Termination Authors:Yoav Alon, Cristina David View a PDF of the paper titled Transformers for Program Termination, by Yoav Alon and 1 other authors View PDF HTML (experimental) Abstract:Determining whether a program terminates is a core challenge in program analysis with direct implications for correctness, verification, and security. We investigate whether transformer architectures can recognise termination patterns directly from source code and how their strengths can be amplified through ensembles. To overcome the extreme scarcity of non-terminating examples, we design an ensemble framework of compact transformer encoders, systematically trained with a suite of imbalance-aware loss functions and class-aware sampling techniques. By combining models trained with distinct loss functions, our ensembles achieve substantially stronger performance than any single transformer, outperforming both powerful off-the-shelf LLMs and graph-based methods. Finally, we introduce an attribution pipeline that produces syntax-aware explanations for the termination estimation. Comments: Subjects: Programming Languages (cs.PL); Machine Learning (cs.LG) Cite as: arXiv:2604.00039 [cs.PL]   (or arXiv:2604.00039v1 [cs.PL] for this version)   https://doi.org/10.48550/arXiv.2604.00039 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Yoav Alon PhD [view e...