Separation of concerns keeps knowledge storage and reasoning in distinct components. In retrieval-augmented generation, a retriever finds information while a generator synthesizes it, improving maintainability, updateability, and comprehensibility. section 4.7
The pattern provides practical architectural benefits:
- Specialized components can collectively outperform a monolithic system when each is effective at its own task. section 4.7
- Independent knowledge storage enables updates and real-time adaptation without retraining the entire system. section 4.7
- Simple retrieve-and-generate pipelines favor deployment and adoption over the complexity of joint training. section 4.7
Table 4.2 Canonical RAG vs naive RAG
|
Feature
|
Canonical RAG
(Lewis et al., 2020)
|
Naive RAG (Industry Standard)
|
|
Definition
|
The original research involved sophisticated joint
training and explicit probabilistic marginalization over retrieved documents.
|
A simplified, practical pattern using off-the-shelf
components, relying on the LLM's inherent capabilities to integrate retrieved
context.
|
|
Training
|
Jointly fine-tuned the retriever and generator for the
specific task.
|
No joint training; uses a pretrained retriever and LLM.
|
|
Document Fusion
|
Employs explicit marginalization (e.g., RAG-Sequence,
RAG-Token) to weigh document contributions.
|
Concatenates retrieved documents into the LLM's prompt;
relies on the LLM's attention to handle context.
|
|
Complexity
|
Higher architectural and computational complexity requires
specialized training.
|
Much simpler to implement and deploy; significantly lower
computational cost.
|
|
Legacy
|
Established the fundamental retrieval-augmented generation
pattern that subsequent work has refined but not replaced.
|
The default architecture for most real-world RAG
applications is due to its accessibility and effectiveness.
|
Compare canonical RAG's jointly trained, explicitly marginalized architecture with naive RAG's off-the-shelf components and prompt-based document fusion. section 4.7
























