Is Knowledge in Multilingual Language Models Cross-Lingually Consistent?
Keywords: Multilingual Models, Fact-checking, Cross-lingual Knowledge Consistency, Self-consistency, Model Parity
TL;DR: We evaluate cross-lingual consistency for factual knowledge by substituting an entity with an equivalent one in other languages that shares the same reference.
Abstract: Few works study the variation and cross-lingual consistency of factual knowledge embedded in multilingual models. However, cross-lingual consistency should be considered to assess cross-lingual transferability, maintain the factuality of the model’s knowledge across languages, and preserve the parity of language model performance. We are thus interested in analyzing, evaluating, and interpreting cross-lingual consistency for factual knowledge. We apply interpretability approaches to analyze a model’s behavior in cross-lingual contexts, discovering that multilingual models show different levels of consistency, subject to either language families or linguistic factors. Further, we identify a cross-lingual consistency bottleneck manifested in middle layers. To mitigate this problem, we try vocabulary expansion, additional cross-lingual objectives, and adding biases from monolingual inputs. We find that all these methods boost cross-lingual consistency to some extent, with cross-lingual supervision offering the best improvement.
Primary Area: applications to computer vision, audio, language, and other modalities
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.
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: 3933
Loading