LatentMT: Machine Translation with Latent Reasoning

2026-07-20

AI Summary

This work addresses the high computational cost and limited performance of traditional machine translation models on low-resource languages, which often rely on large-scale parameters or explicit reasoning chains. The study presents the first systematic exploration of latent inference mechanisms in machine translation, introducing LoopLM—a latent inference recurrent language model that enhances translation quality through recursive computation within hidden states. Leveraging a 2.6B-parameter backbone, LoopLM requires only lightweight fine-tuning and avoids both model scaling and explicit chain-of-thought generation. Evaluated across 32 language directions spanning high-, medium-, and low-resource settings, the method matches or exceeds the performance of models three to five times larger, achieving state-of-the-art results for medium- and low-resource languages while substantially reducing training and inference overhead.

Abstract

Latent-reasoning looped language models (LoopLMs) offer a different scaling path for machine translation (MT): instead of increasing parameter count or emitting explicit chain-of-thought tokens, they spend additional recurrent computation inside hidden states. We introduce LatentMT, the first systematic study of latent-reasoning LoopLMs for machine translation. LatentMT adapts a small 2.6B-parameter backbone model with lightweight training. Across 32 translation directions spanning high-, mid-, and low-resource languages, LatentMT achieves performance comparable to models three to five times larger. It is competitive in a high-resource language and achieves state-of-the-art performance on both mid-resource and low-resource languages. Studying the behavior of scaling the number of recurrent reasoning steps, we find that recurrent computation consistently improves translation quality in early steps, then saturates quickly afterwards. Our mechanistic analysis shows that hidden-representation differences shrink along the recurrent reasoning-step axis, supporting the observed saturation in performance. Finally, our efficiency analysis shows that LatentMT requires lower training and inference compute than much larger non-latent-reasoning models with similar performance, making latent recurrent computation a promising path toward compact, efficient, and strong machine translation.

Problem

• machine translation
• latent reasoning
• recurrent computation
• low-resource languages
• model efficiency

Innovation

• latent reasoning
• looped language models
• machine translation
• recurrent computation
• parameter efficiency