Retrieval-augmented generation

LlamaIndex

Superseded baseline#86 of 1,179 most-superseded

Cited as a baseline — critiqued by newer work, not yet beaten on a benchmark here

3 papers critique it · 0 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites LlamaIndex as a baseline.

While there are some existing RAG toolkits such as LangChain and LlamaIndex, they are typically complex and cumbersome, restricting researchers from tailoring processes to their specific needs.
FlashRAG: A Modular Toolkit for Efficient Retrieval-Augmented Generation Research
Toolkits like LangChain and LlamaIndex, modularize the RAG process, increasing adaptability and broadening its applications. However, they are typically cumbersome, making adaptation to new data challenging and validating or optimizing innovative methods inconvenient.
XRAG: eXamining the Core -- Benchmarking Foundational Components in Advanced Retrieval-Augmented Generation
LlamaIndex...offer generic off-the-shelf application modules for high-level RAG development tailored for production-ready applications. Furthermore, %but do not offer the necessary code flexibility of a research environment.
BERGEN: A Benchmarking Library for Retrieval-Augmented Generation

What to use instead

Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.