Mending the Holes: Mitigating Reward Hacking in Reinforcement Learning for Multilingual Translation
This addresses the challenge of improving machine translation for low-resource languages without relying on scarce parallel data, though it is incremental as it builds on existing reinforcement learning and quality estimation methods.
The paper tackles the problem of reward hacking in reinforcement learning for multilingual translation, particularly for low-resource languages, by introducing WALAR, a method that uses monolingual text to improve translation capabilities; the result is a model that outperforms LLaMAX on 1400 language directions on the Flores-101 dataset.
Large Language Models (LLMs) have demonstrated remarkable capability in machine translation on high-resource language pairs, yet their performance on low-resource translation still lags behind. Existing post-training methods rely heavily on high-quality parallel data, which are often scarce or unavailable for low-resource languages. In this paper, we introduce WALAR, a reinforcement training method using only monolingual text to elevate LLMs' translation capabilities on massive low-resource languages while retaining their performance on high-resource languages. Our key insight is based on the observation of failure modes (or "holes") in existing source-based multilingual quality estimation (QE) models. Reinforcement learning (RL) using these QE models tends to amplify such holes, resulting in poorer multilingual LLMs. We develop techniques including word alignment and language alignment to mitigate such holes in WALAR's reward for RL training. We continually trained an LLM supporting translation of 101 languages using WALAR. The experiments show that our new model outperforms LLaMAX, one of the strongest open-source multilingual LLMs by a large margin on 1400 language directions on Flores-101 dataset.