Attribute-Controlled Translation with Preference Optimization

This repository contains the code for the paper "Attribute-Controlled Translation with Preference Optimization" (Findings of EACL 2026).

Installation

• Python: 3.11.11
• pip install -r requirements.txt
• You can download baseline checkpoints from Huggingface using the scripts/general/download_pretrained_lm.py script into the ./pretrained_lms folder.

Train/eval models

You can find all the scripts in the scripts folder to reproduce the results in the paper.

Citation

@inproceedings{jauregi-unanue-etal-2026-attribute,
    title = "Attribute-Controlled Translation with Preference Optimization",
    author = "Jauregi Unanue, Inigo  and
      Sadoughi, Najmeh  and
      Bhat, Vimal  and
      Liu, Zhu  and
      Piccardi, Massimo",
    editor = "Demberg, Vera  and
      Inui, Kentaro  and
      Marquez, Llu{\'i}s",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EACL} 2026",
    month = mar,
    year = "2026",
    address = "Rabat, Morocco",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.findings-eacl.209/",
    pages = "4031--4057",
    ISBN = "979-8-89176-386-9",
    abstract = "Attribute-controlled translation (ACT) seeks to produce translations that satisfy specific constraints on linguistic and stylistic attributes. While careful prompt engineering can enable large language models to perform strongly in this task, its effectiveness is mainly limited to models of very large size. For this reason, in this paper we set to improve the performance of language models of more contained size by leveraging the contrastive nature of ACT tasks with preference optimization, as well as exploiting knowledge distillation with synthetically-generated training samples from larger models. As a resource for this investigation, we also introduce PREF-FAME-MT, a large, contrastive, formality-controlled parallel corpus which has been generated by expanding the existing FAME-MT dataset with synthetic contrastive samples. Experiments conducted over three datasets for formality- and gender-controlled translation with 71 distinct language pairs have demonstrated the effectiveness of the proposed approach at simultaneously improving attribute matching and translation quality. We release all our code and datasets to allow reproduction and expansion of our work."
}