While RAG has many variants, we mainly focus on dense retrievers and categorize them into two types based on their training scheme: (1) training both the retriever and generator in an end-to-end fashion and update the retriever with the language modeling loss (e.g.

The integration of synthetically generated data in the second and third steps enhances the models' capabilities in interpreting and processing legal texts, effectively reaching state-of-the-art performance and outperforming

Surgical video-language pretraining (VLP) faces unique challenges due to the knowledge domain gap and the scarcity of multi-modal data.

Research on scaling large language models (LLMs) has primarily focused on model parameters and training data size, overlooking the role of vocabulary size.

In particular, we minimize the imitation gap by matching the action-value moments of the teacher's behavior from both on-and off-policy perspectives.

Reconstructing past events requires reasoning across long time horizons.

However, while popular, the system's suitability for assessing entities with constant skill levels, such as LLMs, remains relatively unexplored. We study two fundamental axioms that evaluation methods should adhere to: reliability and transitivity .