作者位次 #3/8ACL 2026Poster· Natural Language Processing
What Does LLM Refinement Actually Improve? A Systematic Study on Document-Level Literary Translation
Shaomu Tan · Dawei Zhu · Ke Tran · Michael Denkowski · Sony Trenous · Leonardo F. R. Ribeiro · Bill Byrne · Felix Hieber
摘要
Iterative refinement is a simple inference-time strategy for machine translation: given an initial translation, an LLM revises it without additional training. Yet document-scale refinement remains poorly understood: 1) which pipelines work best, 2) what quality dimensions improve, and 3) how refi...