Language steering in latent space to mitigate unintended code-switching

Abstract:Multilingual Large Language Models (LLMs) often exhibit hallucinations such as unintended code-switching, reducing reliability in downstream tasks. We propose latent-space language steering, a lightweight inference-time method that identifies language directions via Principal Component Analysis (PCA) on parallel translations and steers token embeddings along these axes to control language identity. Our approach mitigates code-switching (reduced next-token distributional divergence by up to 55\%, reduced Code-Switching Index by 63--99\%) while preserving broad semantic direction with negligible computational overhead and requires only minimal parallel data for calibration. Additionally, we evaluate language classification using a single principal component (95-99\% accuracy) and further analyze the layer-wise evolution of language representations.

Computation and Language (cs.CL); Machine Learning (cs.LG)

arXiv:2510.13849 [cs.CL]

(or arXiv:2510.13849v4 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2510.13849 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.