Language Space

Language space research investigates how words and sentences are represented and manipulated computationally, aiming to understand and improve natural language processing (NLP). Current efforts focus on developing methods to bridge different modalities (e.g., vision and language) using techniques like linear transformations and optimal transport, as well as refining model architectures like transformers to enhance interpretability and robustness. This work is crucial for advancing NLP applications, including improving the safety and privacy of large language models, and enabling more efficient and effective solutions for tasks like machine translation and question answering.