Loong: A Human-Like Long Document Translation Agent with Observe-and-Act Adaptive Context Selection

2026-05-28

AI Summary

This work addresses the limitations of large language models in long-document translation, where constrained context windows hinder global coherence and redundant context degrades output quality. To overcome these challenges, the authors propose Loong, a human-like translation agent that emulates human cognitive processes through a structured 3E memory module—comprising Essence, Exemplar, and Entity components—to retain critical information. Loong integrates deep reasoning with preference-based reinforcement learning to enable adaptive selection of relevant historical context. Experimental results demonstrate that Loong achieves an average improvement of 13.0 BLEU points on English↔Chinese, German, and French long-document translation tasks, while significantly enhancing cross-domain generalization, robustness to noise, and stability in handling extremely long texts.

Abstract

Document-level translation remains one of the most challenging tasks for large language models, which are constrained by limited context windows that impede global cohesion, while simultaneously suffering from redundant contextual information that degrades translation quality. To address this, we propose a human-like long document translation agent called Loong, which leverages a 3E memory module (Essence-Exemplar-Entity) to store summaries, sentence pairs, and entity records as historical context. Instead of passively attending to all history, Loong performs deep reasoning to adaptively identify the optimal context for translation guidance. Loong optimizes its context policy through reinforcement learning, utilizing preference data derived from its own sampled observe-and-act reasoning trajectories. Empirical evaluations demonstrate that Loong achieves substantial translation quality improvements in English $\Leftrightarrow$ Chinese, German, and French directions, with average gains of up to 13.0 points across the three evaluation metrics. Furthermore, Loong exhibits strong generalization across domains and robustness against contextual noise, while maintaining remarkable stability in ultra-long document translation. Our code is released at https://github.com/YutongWang1216/LoongDocMT.

Problem

• document-level translation
• context window limitation
• redundant context
• translation quality
• long document

Innovation

• adaptive context selection
• 3E memory module
• observe-and-act reasoning
• reinforcement learning
• document-level machine translation