{"slug": "augmenting-text-to-increase-translation-difficulty", "title": "Augmenting Text to Increase Translation Difficulty", "summary": "Researchers propose Adversarial Translation Optimization (ATO), a gradient-based method that augments existing machine translation benchmarks to increase difficulty, lowering average translation quality (xCOMET) from 0.93 to 0.82, compared to 0.88 for paraphrasing and 0.86 for a zero-shot baseline. The method uses beam search to iteratively replace tokens, offering an alternative to LLM-based dataset creation without prompting or human curation. The team releases two datasets of 350 English texts each and the code.", "body_md": "arXiv:2608.15932v1 Announce Type: new\nAbstract: 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.", "url": "https://wpnews.pro/news/augmenting-text-to-increase-translation-difficulty", "canonical_source": "https://www.machinebrief.com/news/augmenting-text-to-increase-translation-difficulty-yfld", "published_at": "2026-08-18 04:00:00+00:00", "updated_at": "2026-08-18 05:40:58.582834+00:00", "lang": "en", "topics": ["machine-learning", "natural-language-processing", "ai-research", "ai-tools"], "entities": ["arXiv", "Adversarial Translation Optimization (ATO)", "xCOMET"], "alternates": {"html": "https://wpnews.pro/news/augmenting-text-to-increase-translation-difficulty", "markdown": "https://wpnews.pro/news/augmenting-text-to-increase-translation-difficulty.md", "text": "https://wpnews.pro/news/augmenting-text-to-increase-translation-difficulty.txt", "jsonld": "https://wpnews.pro/news/augmenting-text-to-increase-translation-difficulty.jsonld"}}