Translate Smart, not Hard: Cascaded Translation Systems with Quality-Aware Deferral
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916619312168960 |
|---|---|
| author | Farinhas, António Guerreiro, Nuno M. Agrawal, Sweta Rei, Ricardo Martins, André F. T. |
| author_facet | Farinhas, António Guerreiro, Nuno M. Agrawal, Sweta Rei, Ricardo Martins, André F. T. |
| contents | Larger models often outperform smaller ones but come with high computational costs. Cascading offers a potential solution. By default, it uses smaller models and defers only some instances to larger, more powerful models. However, designing effective deferral rules remains a challenge. In this paper, we propose a simple yet effective approach for machine translation, using existing quality estimation (QE) metrics as deferral rules. We show that QE-based deferral allows a cascaded system to match the performance of a larger model while invoking it for a small fraction (30% to 50%) of the examples, significantly reducing computational costs. We validate this approach through both automatic and human evaluation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_12701 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Translate Smart, not Hard: Cascaded Translation Systems with Quality-Aware Deferral Farinhas, António Guerreiro, Nuno M. Agrawal, Sweta Rei, Ricardo Martins, André F. T. Computation and Language Artificial Intelligence Machine Learning Larger models often outperform smaller ones but come with high computational costs. Cascading offers a potential solution. By default, it uses smaller models and defers only some instances to larger, more powerful models. However, designing effective deferral rules remains a challenge. In this paper, we propose a simple yet effective approach for machine translation, using existing quality estimation (QE) metrics as deferral rules. We show that QE-based deferral allows a cascaded system to match the performance of a larger model while invoking it for a small fraction (30% to 50%) of the examples, significantly reducing computational costs. We validate this approach through both automatic and human evaluation. |
| title | Translate Smart, not Hard: Cascaded Translation Systems with Quality-Aware Deferral |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2502.12701 |