Search papers, labs, and topics across Lattice.
This paper introduces Variable Entropy Policy Optimization (VEPO), a reinforcement learning framework with verifiable rewards designed to improve the performance of LLMs on low-resource languages. VEPO enforces structural constraints like sequence length and linguistic well-formedness during training, using a variable entropy mechanism to balance literal fidelity and semantic naturalness. Experiments across FLORES-200, COMET-22, and chrF datasets show that VEPO significantly enhances tokenization efficiency and translation quality for underrepresented languages.
Low-resource language models can get a major boost in translation quality and tokenization efficiency by using reinforcement learning to directly enforce structural constraints like sequence length and linguistic well-formedness during training.
Large language models frequently exhibit suboptimal performance on low resource languages, primarily due to inefficient subword segmentation and systemic training data imbalances. In this paper, we propose Variable Entropy Policy Optimization (VEPO), which leverages Reinforcement Learning with Verifiable Rewards to incorporate deterministic structural constraints into the policy alignment process. This framework ensures prescribed sequence length, robust format consistency, and rigorous linguistic well formedness, all enforced during training. Central to our approach is a variable entropy mechanism that enables the model to dynamically calibrate the equilibrium between literal fidelity and semantic naturalness by modulating the exploration exploitation manifold. By integrating entropy tempered advantage estimation with asymmetric clipping, VEPO sustains robust exploration while mitigating policy collapse. Empirical evaluations across 90 FLORES-200, COMET-22, chrF directions demonstrate that VEPO yields substantial improvements in both tokenization efficiency and translation quality, bridging the performance gap for underrepresented languages.
