Deciphering the Interplay of Parametric and Non-parametric Memory in Retrieval-augmented Language Models

TL;DR

This paper investigates how retrieval-augmented language models balance parametric knowledge and retrieved information, revealing that models prefer context when both sources are available and identifying mechanisms behind this decision.

Contribution

It introduces a causal mediation analysis approach to disentangle and understand the internal decision processes of RAG models in utilizing parametric versus non-parametric knowledge.

Findings

Models rely more on retrieved context when both sources are available.

Multiple internal mechanisms influence how context relevance is determined.

The study provides insights into the internal computations supporting copying and relevance detection.

Abstract

Generative language models often struggle with specialized or less-discussed knowledge. A potential solution is found in Retrieval-Augmented Generation (RAG) models which act like retrieving information before generating responses. In this study, we explore how the \textsc{Atlas} approach, a RAG model, decides between what it already knows (parametric) and what it retrieves (non-parametric). We use causal mediation analysis and controlled experiments to examine how internal representations influence information processing. Our findings disentangle the effects of parametric knowledge and the retrieved context. They indicate that in cases where the model can choose between both types of information (parametric and non-parametric), it relies more on the context than the parametric knowledge. Furthermore, the analysis investigates the computations involved in \emph{how} the model uses the information from the context. We find that multiple mechanisms are active within the model and can be detected with mediation analysis: first, the decision of \emph{whether the context is relevant}, and second, how the encoder computes output representations to support copying when relevant.