Augmenting Text to Increase Translation Difficulty

2026-08-16

AI Summary

This study addresses the saturation of machine translation benchmarks and insufficient model discrimination by proposing an adversarial translation optimization framework. The method innovatively integrates a differentiable difficulty estimator with beam search to iteratively optimize token replacements via gradient-based updates, enabling automatic construction of challenging evaluation data without reliance on large language models or manual annotation. Experiments demonstrate that the generated samples achieve an xCOMET score of 0.82, indicating significantly higher difficulty than existing baselines while maintaining grammatical validity. This work establishes an efficient and scalable paradigm for machine translation evaluation, with both code and datasets publicly released to facilitate further research.

Abstract

As state-of-the-art machine translation models saturate standard benchmarks, the field needs more challenging evaluations to distinguish between models of varying quality. We propose augmenting existing benchmarks to increase translation difficulty by combining adversarial optimization with a differentiable translation difficulty estimator. Our Adversarial Translation Optimization (ATO) uses gradients from a combined difficulty and fluency objective to iteratively replace tokens. Because each step branches over candidate substitutions at every position, optimization becomes a tree search problem, which we address with Beam Search. ATO offers a gradient-based alternative to LLM-based dataset creation without LLM prompting, expensive human curation, or task-specific model training. Our ATO-modified benchmark lowers average translation quality (xCOMET) from 0.93 to 0.82, compared to 0.88 for paraphrasing and 0.86 for a zero-shot baseline. Human evaluation shows the modified texts are somewhat less natural than the baselines but remain reasonably grammatical and plausible while being substantially harder to translate. We release two datasets of 350 English texts each, generated by our methods, as well as the code.

Problem

• Machine Translation Evaluation
• Benchmark Saturation
• Translation Difficulty
• Model Discrimination

Innovation

• Adversarial Translation Optimization
• Differentiable Difficulty Estimator
• Gradient-based Token Substitution
• Beam Search
• Translation Benchmark Augmentation