SAMAS: A Spectrum-Guided Multi-Agent System for Achieving Style Fidelity in Literary Translation
This addresses the challenge of style preservation in literary translation for translators and literary applications, representing a novel approach rather than an incremental improvement.
The paper tackles the problem of preserving an author's unique literary style in translations, which current LLMs often fail to do, by introducing SAMAS, a framework that quantifies style into a spectrum and dynamically assembles specialized agents, achieving statistically significant improvements in style fidelity while maintaining competitive semantic accuracy.
Modern large language models (LLMs) excel at generating fluent and faithful translations. However, they struggle to preserve an author's unique literary style, often producing semantically correct but generic outputs. This limitation stems from the inability of current single-model and static multi-agent systems to perceive and adapt to stylistic variations. To address this, we introduce the Style-Adaptive Multi-Agent System (SAMAS), a novel framework that treats style preservation as a signal processing task. Specifically, our method quantifies literary style into a Stylistic Feature Spectrum (SFS) using the wavelet packet transform. This SFS serves as a control signal to dynamically assemble a tailored workflow of specialized translation agents based on the source text's structural patterns. Extensive experiments on translation benchmarks show that SAMAS achieves competitive semantic accuracy against strong baselines, primarily by leveraging its statistically significant advantage in style fidelity.