Mending the Holes: Mitigating Reward Hacking in Reinforcement Learning for Multilingual LLMs

ACL ARR 2026 January Submission7863 Authors

Keywords: Machine Translation, Large Language Model, Reinforcement Learning, Reward Design

Abstract: 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.

Paper Type: Long

Research Area: Machine Translation

Research Area Keywords: multilingual MT, reinforcement learning

Contribution Types: Approaches to low-resource settings

Languages Studied: 102 languages coverd by Flores-101

Submission Number: 7863