Explanation Dataset

Explanation datasets are collections of data designed to improve the interpretability and trustworthiness of machine learning models, particularly large language models (LLMs) and vision transformers (ViTs). Current research focuses on creating datasets with diverse sources, rich annotations (including explanations generated by LLMs and human-curated labels), and multi-level explanations bridging image/patch, sentence, and dataset-level interpretations. These datasets are crucial for evaluating and improving model performance, addressing issues like bias, hallucination, and generalization, and ultimately fostering more reliable and explainable AI systems across various applications.