Abstract Reinforcement learning (RL) has been proven to be an effective and robust method for training neural machine translation systems, especially when paired with powerful reward models that accurately assess translation quality. However, most research has focused on RL methods that use sentence-level feedback, leading to inefficient learning signals due to the reward sparsity problem—the model receives a single score for the entire sentence. To address this, we propose a novel approach that leverages fine-grained, token-level quality assessments along with error severity levels using RL methods. Specifically, we use xCOMET, a state-of-the-art quality estimation system, as our token-level reward model. We conduct experiments on small and large translation datasets with standard encoder-decoder and large language models-based machine translation systems, comparing the impact of sentence-level versus fine-grained reward signals on translation quality. Our results show that training with token-level rewards improves translation quality across language pairs over baselines according to both automatic and human evaluation. Furthermore, token-level reward optimization improves training stability, evidenced by a steady increase in mean rewards over training epochs.

In this paper, disruptive research using generative diffusion models (DMs) with an attention-based encoder-decoder backbone is conducted to automate the sizing of analog integrated circuits (ICs). Unlike time-consuming optimization-based methods, the encoder-decoder DM is able to sample accurate solutions at push-button speed by solving the inverse sizing problem. Experimental results show that the proposed model outperforms the most recent deep learningbased techniques, presenting higher generalization capabilities to performance targets not seen during training.