Making Retrieval-Augmented Language Models Robust to Irrelevant Context

By Ori Yoran, Tomer Wolfson, Ori Ram, Jonathan Berant. International Conference on Learning Representations, 2023

405 citations recorded by the source corpus.

Retrieval-augmented language models (RALMs) hold promise to produce language understanding systems that are are factual, efficient, and up-to-date. An important desideratum of RALMs, is that retrieved information helps model performance when it is relevant, and does not harm performance when it is not. This is particularly important in multi-hop reasoning scenarios, where misuse of irrelevant evidence can lead to cascading errors. However, recent work has shown that retrieval augmentation can sometimes have a negative effect on performance. In this work, we present a thorough analysis on five open-domain question answering benchmarks, characterizing cases when retrieval reduces accuracy. We then propose two methods to mitigate this issue. First, a simple baseline that filters out retrieved passages that do not entail question-answer pairs according to a natural language inference (NLI) model. This is effective in preventing performance reduction, but at a cost of also discarding relevant passages. Thus, we propose a method for automatically generating data to fine-tune the language model to properly leverage retrieved passages, using a mix of relevant and irrelevant contexts at training time. We empirically show that even 1,000 examples suffice to train the model to be robust to irrelevant contexts while maintaining high performance on examples with relevant ones.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/3aee33831e0bdea1a1eaae21c7586e4f7c0396d6

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Search-o1: Agentic Search-Enhanced Large Reasoning Models (2025, neutral): “…redundant information, directly inputting them into LRMs may disrupt the original coherence of reasoning and even introduce noise [62, 72, 26]. disrupt the original coherence of reasoning introduce noise (Wu et al., 2024; Yoran 2 Related Work (2) Limited Ability to Understand Long Documents.”
Retrieval-Augmented Generation for Large Language Models: A Survey (2023, background): “Furthermore, such singular retrieval is typically insufficient for complex problems demanding multi-step reasoning, as it provides a limited scope of information [ Yoran et al. , 2023 ] .”
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (2023, background): “…tokens, and the bold text indicates the most desirable critique tokens. x, y, d indicate input, output, and a relevant passage, respectively. of retrieved passages prepended to input, or pre-train a retriever and LM jointly, followed by few-shot fine-tuning on task datasets (Izacard et al., 2022b).”
A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models (2024, neutral): “To provide a clear overview of the applications of RA-LLMs, we will review them from three perspectives: NLP applications , downstream tasks , and domain-specific applications .”
Small Encoders Can Rival Large Decoders in Detecting Groundedness (2025, unknown): “However, their tendency to hallucinate when the query lacks support from the provided context or when the model faces noisy retrieval [Yoran et al., 2024, Wu et al., 2024] raises concerns regarding their reliability [Ji et al., 2023].”
Out of Style: RAG's Fragility to Linguistic Variation (2025, unknown): “Moreover, unlike standalone LLMs, RAG systems incorporate multiple interdependent components, making them susceptible to cascading errors arising at both the retrieval and generation stages (Asai et al., 2023; Yoran et al., 2024a; Kim et al., 2025).”
A Survey of Model Architectures in Information Retrieval (2025, unknown): “We omit detailed discussions here and refer readers to the original papers.”
mFollowIR: a Multilingual Benchmark for Instruction Following in Retrieval (2025, unknown): “Another reason to use a reranking task 3 is the benefits it brings to a RAG system, where irrelevant documents placed near the top of the ranked list will make it easier for the LM to be distracted and answer incorrectly [43, 49, 61].”