When Flores Bloomz Wrong: Cross-Direction Contamination in Machine Translation Evaluation
Abstract
Large language models (LLMs) can be benchmark-contaminated, resulting in inflated scores that mask memorization as generalization, and in multilingual settings, this memorization can even transfer to "uncontaminated" languages. Using the FLORES-200 translation benchmark as a diagnostic, we study two 7-8B instruction-tuned multilingual LLMs: Bloomz, which was trained on FLORES, and Llama as an uncontaminated control. We confirm Bloomz's FLORES contamination and demonstrate that machine translation contamination can be cross-directional, artificially boosting performance in unseen translation directions due to target-side memorization. Further analysis shows that recall of memorized references often persists despite various source-side perturbation efforts like paraphrasing and named entity replacement. However, replacing named entities leads to a consistent decrease in BLEU, suggesting an effective probing method for memorization in contaminated models.
Cite
@article{arxiv.2601.20858,
title = {When Flores Bloomz Wrong: Cross-Direction Contamination in Machine Translation Evaluation},
author = {David Tan and Pinzhen Chen and Josef van Genabith and Koel Dutta Chowdhury},
journal= {arXiv preprint arXiv:2601.20858},
year = {2026}
}
Comments
5 pages of content, 15 total. 5 figures, 12 tables total. Accepted to EACL 2026 main conference. Code can be found here: github.com/Mr-Ao-25/cross-ling-contamination