Redefining Machine Translation on Social Network Services with Large Language Models

TL;DR

This work addresses the challenge of translating culturally rich SNS content by introducing RedTrans, a 72B model specialized for SNS translation. It pairs RedTrans with RedTrans-Bench, the first large-scale benchmark for SNS translation, and a training pipeline that combines Dual-LLM Back-Translation Sampling for diverse SFT data with Rewritten Preference Optimization (RePO) to improve RLHF reliability. Empirical results show RedTrans outperforms state-of-the-art LLMs on SNS-relevant benchmarks and maintains strong performance on general MT tasks, with a real-world deployment validating domain-specific adaptation. The approach demonstrates that tailoring training data generation and preference alignment to SNS cultural nuances markedly bridges the gap between generic translation models and culturally grounded, user-aligned translations.

Abstract

The globalization of social interactions has heightened the need for machine translation (MT) on Social Network Services (SNS), yet traditional models struggle with culturally nuanced content like memes, slang, and pop culture references. While large language models (LLMs) have advanced general-purpose translation, their performance on SNS-specific content remains limited due to insufficient specialized training data and evaluation benchmarks. This paper introduces RedTrans, a 72B LLM tailored for SNS translation, trained on a novel dataset developed through three innovations: (1) Supervised Finetuning with Dual-LLM Back-Translation Sampling, an unsupervised sampling method using LLM-based back-translation to select diverse data for large-scale finetuning; (2) Rewritten Preference Optimization (RePO), an algorithm that identifies and corrects erroneous preference pairs through expert annotation, building reliable preference corpora; and (3) RedTrans-Bench, the first benchmark for SNS translation, evaluating phenomena like humor localization, emoji semantics, and meme adaptation. Experiments show RedTrans outperforms state-of-the-art LLMs. Besides, RedTrans has already been deployed in a real-world production environment, demonstrating that domain-specific adaptation, effectively bridges the gap between generic and culturally grounded translation systems.

Figures (7)

• Figure 1: Overall Framework. We enhance translation models by leveraging open-source corpora and high-engagement social media content. To ensure quality, we employ back-translation sampling and preference optimization techniques. For comprehensive evaluation, we introduce RedTrans-Bench.
• Figure 2: Overview of RedTrans-Bench dataset characteristics.
• Figure 3: The Chinese word cloud of RedTrans-Bench.
• Figure 4: Top 50 Chinese Verb-Noun structures in RedTrans-Bench instructions.
• Figure 5: The reward margin between chosen and rejected responses shows a steady upward trend throughout training, with increased volatility and higher peaks in later stages.