RAAT (Retrieval-augmented generation): superseded — cited as a baseline and beaten by newer methods. 2 paper(s) critique it, 3 beat it on benchmarks — #41 of 1179 most-superseded. Sub-problem: cluster led by RetRobust. Newer alternatives in the same sub-problem include Stable-RAG, Neuro-RIT, ERA, TTARAG, DMA (Dynamic Memory Alignment).

Method Drift›Retrieval-augmented generation

Superseded baseline#41 of 1,179 most-superseded

RAAT

RAAT: Relation-Augmented Attention Transformer for Relation Modeling in Document-Level Event Extraction

Retrieval-augmented generation · first seen Jun 7, 2022

superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 3 beat it on benchmarks

What papers say

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

“It is noted that the model trained with the RAAT method performs worse than the baseline model. This is because its training data follows the setting of {d_golden d_noise}. While this setting trains the model to be robust against noise, the inclusion of golden documents in all data deviates from the conventional RAG setup, resulting in poorer performance under the standard setting.”
— DACL-RAG: Data Augmentation Strategy with Curriculum Learning for Retrieval-Augmented Generation
“However, these robust training approaches are primarily applied to small or weak LMs with fewer than 7 billion parameters. Thus, there's an urgent need to explore whether complex robust training is still necessary to improve the robustness and generalization of bigger or stronger models when dealing with noisy contexts.”
— On the Diminishing Returns of Complex Robust RAG Training in the Era of Powerful LLMs

Beaten on benchmarks

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

TTARAG beats RAAT · Overall [Llama-2-7b-chat]
30.5 vs 22.7
Predict the Retrieval! Test time adaptation for Retrieval Augmented Generation
DTA beats RAAT · Acc [Llama-2-7b]
64.1 vs 46.2
Divide-Then-Align: Honest Alignment based on the Knowledge Boundary of RAG
Stable-RAG beats RAAT · SubEM Average [LLaMA3-8B-Instruct, Contriever]
52.34 vs 47.50
Stable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented Generation
Stable-RAG beats RAAT · SubEM Average [LLaMA3-8B-Instruct, DPR]
52.34 vs 47.50
Stable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented Generation
Stable-RAG beats RAAT · SubEM Average [Qwen3-8B, Contriever]
50.27 vs 47.94
Stable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented Generation
Stable-RAG beats RAAT · SubEM Average [Qwen3-8B, DPR]
50.27 vs 47.94
Stable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented Generation

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.