Scholar Feed

Retrieval-augmented generation

ReliabilityRAG

ReliabilityRAG: Effective and Provably Robust Defense for RAG-based Web-Search

Superseded baseline#43 of 1,179 most-superseded · first seen Sep 27, 2025

Superseded — cited as a baseline and beaten by newer methods

3 papers critique it · 2 beat it on benchmarks

What papers say

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

However, these methods introduce significant runtime overhead, as the retrieve-and-filter process must be performed for every user question, and they often achieve limited effectiveness and/or degrade the utility of the RAG system
— CleanBase: Detecting Malicious Documents in RAG Knowledge Databases
“However, as this problem is NP-hard, processing large scale retrieval sets requires discarding a substantial number of documents to reduce computational costs, leading to significant information loss.”
— BiRD: A Bidirectional Ranking Defense Mechanism for Retrieval Augmented Generation
“Heuristic aggregation or filtering~xiang2024certifiably often causes utility loss, while optimization-based consistency selection~shenreliabilityrag typically relies on approximations without strong guarantees.”
— RADAR: Defending RAG Dynamically against Retrieval Corruption

Beaten on benchmarks

Head-to-head results where a newer method reports beating ReliabilityRAG. Values are copied from the source paper's tables — verify against the cited paper.

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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.