When Flores Bloomz Wrong: Cross-Direction Contamination in Machine Translation Evaluation
This work addresses the problem of inflated evaluation scores due to benchmark contamination for researchers and practitioners in machine translation, highlighting a critical issue in model assessment.
The study investigated cross-direction contamination in machine translation evaluation, showing that models trained on benchmarks like FLORES-200 can artificially boost performance in unseen translation directions due to memorization, with named entity replacement effectively probing this issue and causing a consistent BLEU score decrease.
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.