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.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2603.12983 [cs.CL] |
| (or arXiv:2603.12983v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2603.12983 arXiv-issued DOI via DataCite |
From: Boxuan Lyu [view email]
[v1]
Fri, 13 Mar 2026 13:34:45 UTC (105 KB)
[v2]
Mon, 16 Mar 2026 10:15:07 UTC (105 KB)
[v3]
Mon, 25 May 2026 06:39:08 UTC (108 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。