Work

Is Human Annotation Necessary? Iterative MBR Distillation for Error Span Detection in Machine Translation

Year: 2026

Type: preprint

Abstract: Error Span Detection (ESD) is a crucial subtask in Machine Translation (MT) evaluation, aiming to identify the location and severity of translation errors. While fine-tuning models on human-annotated data improves ESD performance, acquiring such data is expensive and prone to inconsistencies among annotators. To address this, we propose a novel self-evolution framework based on Minimum Bayes Risk (MBR) decoding, named Iterative MBR Distillation for ESD, which eliminates the reliance on human annotations by leveraging an off-the-shelf LLM to generate pseudo-labels. Extensive experiments on the WMT Metrics Shared Task datasets demonstrate that models trained solely on these self-generated pseudo-labels outperform both unadapted base model and supervised baselines trained on human annotations at the system and span levels, while maintaining competitive sentence-level performance.

Source: arXiv (Cornell University)

DOI: 10.48550/arxiv.2603.12983

Authors: Lyu, Boxuan, Song, Haiyue, Qu, Zhi

Institutions: Institute of Science Tokyo, National Institute of Information and Communications Technology

Topic: Natural Language Processing Techniques

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funders: Iran Telecommunication Research Center, Japan Society for the Promotion of Science, National Institute of Information and Communications Technology

Award: Improving Large Language Models for Low-Resource Languages with Diverse Writing Systems

OpenAlex ID: W7136189722

Cites:

Cited by:

Related to: