With the rapid advancement of information access systems, evaluating such systems presents both significant challenges and opportunities. In this talk, we explore the forefront of evaluation methodologies tailored for the generative era, emphasizing the multifaceted aspects of assessment and the urgent need for robust, fair, and accurate evaluation strategies that keep pace with technological advancements. Through this presentation, we aim to shed light on the process of quantitatively comparing the output of one Large Language Model (LLM) with another, as well as with the output of a retrieval system. We will explore methods to identify kernels of truth, aiming to provide an environment where correct information is available, identifiable, and accessible.