We are pleased to announce that our paper "Rethinking Pivot Programming Languages in Code Language Models" has been accepted in Empirical Methods in Natural Language Processing as a main conference paper.

The paper investigates whether certain programming languages play a central role in multilingual code language models. Analyzing three models, we find that no single language serves as a universal pivot: Python aligns particularly well with English, while different languages facilitate different types of cross-language transfer.

This paper has been a cooperation between Lukas Galke Poech at the University of South Denmark, and Andor Diera and Matthias Tichy of Ulm University.

The preprint of the paper can be found at https://arxiv.org/abs/2609.22988