Retrieval-augmented generation
Ret-Robust
Superseded baseline#73 of 1,179 most-superseded
Superseded — cited as a baseline and beaten by newer methods
1 papers critique it · 2 beat it on benchmarks
What papers say
Verbatim critique sentences, each from a paper that cites Ret-Robust as a baseline.
Ret-Robust~yoran2023making trains a model to be robust to irrelevant context, with an NLI-based entailment to determine relevance, and only uses the relevance scores to influence the training mixture of relevant vs. irrelevant documents.
Beaten on benchmarks
Head-to-head results where a newer method reports beating Ret-Robust. Values are copied from the source paper's tables — verify against the cited paper.
DTA beats Ret-Robust
64.8 vs 51.6
Acc · [Llama-2-13b]
Divide-Then-Align: Honest Alignment based on the Knowledge Boundary of RAGTTARAG beats Ret-Robust
30.5 vs 26.1
Overall · [Llama-2-7b-chat]
Predict the Retrieval! Test time adaptation 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Recent methods in the same sub-problem, not yet superseded in the knowledge base — arXiv benchmark leaders, not vetted production recommendations.
- Stable-RAGStable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented GenerationApr 21, 2026
- Apr 2, 2026
- Feb 24, 2026
- Jan 16, 2026
- Nov 6, 2025