From RAGs to rich parameters: Probing how language models utilize external knowledge over parametric information for factual queries

TL;DR

This paper investigates how retrieval-augmented language models primarily rely on external context rather than their internal parameters for answering factual questions, revealing a shortcut bias in their reasoning process.

Contribution

It provides mechanistic insights into RAG models, showing they favor context over parametric memory, and introduces analysis methods to understand this behavior.

Findings

Models rely more on context than internal parameters.

Parametric memory is minimally used in answering questions.

Models exhibit shortcut behavior across different architectures.

Abstract

Retrieval Augmented Generation (RAG) enriches the ability of language models to reason using external context to augment responses for a given user prompt. This approach has risen in popularity due to practical applications in various applications of language models in search, question/answering, and chat-bots. However, the exact nature of how this approach works isn't clearly understood. In this paper, we mechanistically examine the RAG pipeline to highlight that language models take shortcut and have a strong bias towards utilizing only the context information to answer the question, while relying minimally on their parametric memory. We probe this mechanistic behavior in language models with: (i) Causal Mediation Analysis to show that the parametric memory is minimally utilized when answering a question and (ii) Attention Contributions and Knockouts to show that the last token residual stream do not get enriched from the subject token in the question, but gets enriched from other informative tokens in the context. We find this pronounced shortcut behaviour true across both LLaMa and Phi family of models.