Large Language Models (LLMs) have showcased exceptional capabilities in various
domains, attracting significant interest from both academia and industry. Despite their …

Speaker voices exhibit acoustic variation. Our preliminary experiments reveal that
normalized voice can significantly improve end-to-end speech translation. To mitigate the …

Knowledge Distillation (KD) compresses computationally expensive pre-trained language
models (PLMs) by transferring their knowledge to smaller models, allowing their use in …

Large language models (LLMs) exhibit varying strengths and weaknesses across different
tasks, prompting recent studies to explore the benefits of ensembling models to leverage …

Data often arrives in sequence over time in real-world deep learning applications such as
autonomous driving. When new training data is available, training the model from scratch …

As large language models (LLMs) continue to advance, ensuring their alignment with
human values becomes increasingly critical. Traditional alignment methods heavily rely on …

We consider scenarios where a very accurate (often small) predictive model using restricted
features is available when training a full-featured (often larger) model. This restricted model …

This thesis investigates the incorporation of knowledge into generative open-domain
dialogue systems, a pivotal challenge in natural language processing characterized by the …