Large Language Models (LLMs) have revolutionized how we interact with information, but they sometimes struggle with accuracy. Retrieval-Augmented Generation (RAG) offers a solution by giving LLMs access to external knowledge. However, evaluating RAG systems has been a challenge due to fragmented research setups. Researchers at NAVER LABS Europe introduce BERGEN, a Python library designed to standardize and streamline RAG experimentation. BERGEN acts as a central hub, bringing together various components essential for RAG, including retrievers, rerankers, LLMs, datasets, and evaluation metrics. This makes it easier for researchers to compare results and build upon each other's work. One of BERGEN's key strengths lies in its use of the Hugging Face Hub. This allows researchers to easily integrate existing resources and add new models and datasets with minimal effort. The team behind BERGEN conducted extensive experiments, benchmarking different RAG configurations and analyzing popular evaluation metrics. They found that retrieval quality plays a critical role in the accuracy and effectiveness of LLM responses. Furthermore, their research emphasizes the importance of re-ranking retrieved information to refine the context provided to the LLM. This extra step greatly enhances the quality of generated answers. BERGEN also sheds light on the limitations of existing benchmarks and the potential need for new datasets tailored to RAG evaluation. Interestingly, their findings suggest that LLMs of all sizes, not just the largest ones, can benefit from retrieval augmentation. BERGEN isn't just for English-language tasks. It supports multilingual datasets, paving the way for broader RAG development and research across different languages. By standardizing the experimental process, BERGEN allows for true apples-to-apples comparisons of RAG approaches and promotes faster advancements in the field. This open-source library is a valuable contribution to the growing world of retrieval-augmented generation, enabling more transparent, reproducible, and collaborative research.