Retrieval-Augmented Generation and Hallucination Reduction

In short

The clearest direct comparisons favor retrieval-augmented generation (RAG) over fine-tuning alone for factual question answering, especially for new or less-popular knowledge, but these studies establish answer-performance advantages rather than a uniform, directly measured reduction in hallucination. [16][18] Separate experiments show that retrieval can reduce hallucinations relative to models without retrieval, although that comparison does not by itself establish superiority over fine-tuning alone. [1][2] The benefits depend on retrieving relevant and sufficient evidence, using it faithfully, and managing conflicts or irrelevant context; simply adding documents does not guarantee grounded answers. [7][8][11][20] Fine-tuning and retrieval are not necessarily alternatives: targeted training can improve evidence use, and a published healthcare scoping review reports advantages for hybrids, while other experiments find that particular fine-tuning recipes worsen performance or encourage hallucination. [8][13][17][27] Overall, the evidence supports a conditional advantage for RAG as a factual-knowledge access strategy, not a universal safety guarantee, and many of the studies explaining its conditions and failure modes are preprints rather than peer-reviewed publications. [7][8][10][11][20][25][26]

This summary rests on the findings below, each with the sentences that support it.

Approaches compared

Where the papers agree

• The direct comparative evidence favors RAG over fine-tuning as a method for accessing factual knowledge, particularly when facts are new or uncommon. [16][18]
  2 supporting sentences

  “Notably, when comparing knowledge injection methods, RAG significantly outperforms FT.” [16] · full text

  “Our findings reveal that while unsupervised fine-tuning offers some improvement, RAG consistently outperforms it, both for existing knowledge encountered during training and entirely new knowledge.” [18]

• Retrieval is not sufficient for faithful generation: relevant evidence can still yield unsupported claims, and irrelevant context can reduce performance. [8][11][20]
  3 supporting sentences

  “In this work, we provide a thorough analysis showing current RALMs are not robust to irrelevant retrieved context, causing them to perform worse on certain tasks.” [8] · full text

  “We find that while LMs can leverage relevant in-context documents, the generated answer is only partially attributable towards the documents, especially for LMs trained without retrieval augmentation.” [11]

  “However, LLMs can still produce information that is unsupported or contradictory to the retrieved contexts.” [20]

• Training can complement retrieval when it targets context use or domain adaptation, so evidence favoring RAG over factual memorization through fine-tuning does not imply that all fine-tuning is counterproductive. [8][27]
  2 supporting sentences

  “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.” [8]

  “FT + RAG systems consistently outperformed FT-only or RAG-only approaches across QA, clinical summarization, report generation, and decision support tasks.” [27]

What they leave open

• A standardized quantitative answer about hallucination reduction relative to fine-tuning alone remains unavailable from these materials because the direct comparisons and the hallucination studies use different comparators, adaptation methods, and outcome measures. [2][13][16][18]
  No paper states it in these words: inferred from what the papers do not study. Check it.
• Faithfulness evaluation does not settle source truthfulness or conflicting-source resolution, leaving an important limit on what grounding alone can establish. [20]
  1 supporting sentence

  “In particular, source documents that contain conflicting information, or misinformation, present a challenge for failure detection due to its ambiguity.” [20] · full text

• Generalization beyond English question answering remains incomplete, despite individual domain-specific and Indonesian detection studies. [20][28]
  2 supporting sentences

  “The bulk of datasets used in HaluBench is in English, which presents a limitation in real-world applications that include multilingual inputs and contexts.” [20] · full text

  “First, most hallucination detection studies focus on English language settings.” [28] · full text

• Lower measured hallucination does not establish preserved usefulness or deployment readiness, especially when repair removes content or healthcare evaluations lack standardization. [25][27]
  2 supporting sentences

  “The strategies occupy different points on a grounding preservation trade-off rather than forming a quality ranking, and choosing between them needs evidence about answer usefulness that automatic metrics cannot supply.” [25]

  “Future work should prioritize standardized evaluation, workflow integration, and governance to enable safe deployment.” [27]

References

1. [1] Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

   Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni et al. · 2020 · arXiv (Cornell University) preprint · cited 6223 times · doi:10.48550/arxiv.2005.11401

   Founding Preprint retrieval-augmented generation evaluation · experiment

   Why it is here: It establishes RAG as a knowledge-access approach and evaluates knowledge-intensive tasks, but the abstract does not make a direct fine-tuning-only comparison.

   Abstract · conclusions of the full text read too

   Large pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limited, and hence on knowledge-intensive tasks, their performance lags behind task-specific architectures. Additionally, providing provenance for their decisions and updating their world knowledge remain open research problems. Pre-trained models with a differentiable access mechanism to explicit non-parametric memory can overcome this issue, but have so far been only investigated for extractive downstream tasks. We explore a general-purpose fine-tuning recipe for retrieval-augmented generation (RAG) -- models which combine pre-trained parametric and non-parametric memory for language generation. We introduce RAG models where the parametric memory is a pre-trained seq2seq model and the non-parametric memory is a dense vector index of Wikipedia, accessed with a pre-trained neural retriever. We compare two RAG formulations, one which conditions on the same retrieved passages across the whole generated sequence, the other can use different passages per token. We fine-tune and evaluate our models on a wide range of knowledge-intensive NLP tasks and set the state-of-the-art on three open domain QA tasks, outperforming parametric seq2seq models and task-specific retrieve-and-extract architectures. For language generation tasks, we find that RAG models generate more specific, diverse and factual language than a state-of-the-art parametric-only seq2seq baseline.

2. [2] Retrieval Augmentation Reduces Hallucination in Conversation

   Kurt Shuster, Spencer Poff, Moya Chen, Douwe Kiela et al. · 2021 · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · cited 739 times · doi:10.18653/v1/2021.findings-emnlp.320

   Founding Retrieval-grounded conversational generation · experiment

   Why it is here: Shows retrieval can improve knowledge-grounded dialogue, but does not compare against fine-tuning.

   Abstract

   Despite showing increasingly human-like conversational abilities, state-of-the-art dialogue models often suffer from factual incorrectness and hallucination of knowledge (Roller et al., 2021).In this work we explore the use of neural-retrieval-in-the-loop architectures -recently shown to be effective in open-domain QA (Lewis et al., 2020b; Izacard and Grave, 2021b) -for knowledge-grounded dialogue, a task that is arguably more challenging as it requires querying based on complex multi-turn dialogue context and generating conversationally coherent responses.We study various types of architectures with multiple components -retrievers, rankers, and encoder-decoders -with the goal of maximizing knowledgeability while retaining conversational ability.We demonstrate that our best models obtain state-of-the-art performance on two knowledge-grounded conversational tasks.The models exhibit open-domain conversational capabilities, generalize effectively to scenarios not within the training data, and, as verified by human evaluations, substantially reduce the well-known problem of knowledge hallucination in state-of-the-art chatbots. * Equal ContributionThe following is a conversation with an AI assistant.The assistant is helpful, creative, clever, and very friendly.Human: Hello, who are you?AI: I am an AI created by OpenAI.How can I help you today?Human: Tell me about Kyunghyun

3. [3] Atlas: Few-shot Learning with Retrieval Augmented Language Models

   Gautier Izacard, Patrick Lewis, María Lomelí, Lucas Hosseini et al. · 2022 · arXiv (Cornell University) preprint · cited 744 times · doi:10.48550/arxiv.2208.03299

   Preprint few-shot retrieval-augmented modeling · experiment

   Why it is here: It evaluates retrieval-augmented models across knowledge-intensive tasks and model settings, but does not compare against fine-tuning alone.

   Abstract · conclusions of the full text read too

   Large language models have shown impressive few-shot results on a wide range of tasks. However, when knowledge is key for such results, as is the case for tasks such as question answering and fact checking, massive parameter counts to store knowledge seem to be needed. Retrieval augmented models are known to excel at knowledge intensive tasks without the need for as many parameters, but it is unclear whether they work in few-shot settings. In this work we present Atlas, a carefully designed and pre-trained retrieval augmented language model able to learn knowledge intensive tasks with very few training examples. We perform evaluations on a wide range of tasks, including MMLU, KILT and NaturalQuestions, and study the impact of the content of the document index, showing that it can easily be updated. Notably, Atlas reaches over 42% accuracy on Natural Questions using only 64 examples, outperforming a 540B parameters model by 3% despite having 50x fewer parameters.

4. [4] Chain of Natural Language Inference for Reducing Large Language Model Ungrounded Hallucinations

   Deren Lei, Yaxi Li, Mengya Hu, Mingyu Wang et al. · 2023 · arXiv (Cornell University) preprint · cited 61 times · doi:10.48550/arxiv.2310.03951

   Preprint Inference-based detection and editing · experiment

   Why it is here: Assesses context-grounded hallucination detection and correction, but does not compare with fine-tuning.

   Abstract · conclusions of the full text read too

   Large language models (LLMs) can generate fluent natural language texts when given relevant documents as background context. This ability has attracted considerable interest in developing industry applications of LLMs. However, LLMs are prone to generate hallucinations that are not supported by the provided sources. In this paper, we propose a hierarchical framework to detect and mitigate such ungrounded hallucination. Our framework uses Chain of Natural Language Inference (CoNLI) for hallucination detection and hallucination reduction via post-editing. Our approach achieves state-of-the-art performance on hallucination detection and enhances text quality through rewrite, using LLMs without any fine-tuning or domain-specific prompt engineering. We show that this simple plug-and-play framework can serve as an effective choice for hallucination detection and reduction, achieving competitive performance across various contexts.

5. [5] HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Models

   Junyi Li, Xiaoxue Cheng, Xin Zhao, Jian‐Yun Nie et al. · 2023 · Conference on Empirical Methods in Natural Language Processing (EMNLP) · cited 749 times · doi:10.18653/v1/2023.emnlp-main.397

   Human-annotated hallucination benchmark · method

   Why it is here: Introduces a benchmark for evaluating hallucination recognition, useful for assessment but not a RAG-versus-fine-tuning comparison.

   Abstract

   Large language models (LLMs), such as Chat-GPT, are prone to generate hallucinations, i.e., content that conflicts with the source or cannot be verified by the factual knowledge.To understand what types of content and to which extent LLMs are apt to hallucinate, we introduce the Hallucination Evaluation benchmark for Large Language Models (HaluEval), a large collection of generated and human-annotated hallucinated samples for evaluating the performance of LLMs in recognizing hallucination.To generate these samples automatically, we propose a two-stage framework, i.e., samplingthen-filtering.Besides, we hire some human labelers to annotate the hallucinations in Chat-GPT responses.The empirical results suggest that ChatGPT is likely to generate hallucinated content related to specific topics by fabricating unverifiable information (i.e., about 19.5% responses).Moreover, existing LLMs face great challenges in recognizing the hallucinations in texts.However, our experiments also prove that providing external knowledge or adding reasoning steps can help LLMs recognize hallucinations.Our benchmark can be accessed at https://github.com/RUCAIBox/HaluEval.

6. [6] In-Context Retrieval-Augmented Language Models

   Ori Ram, Yoav Levine, Itay Dalmedigos, Dor Muhlgay et al. · 2023 · Transactions of the Association for Computational Linguistics · cited 752 times · doi:10.1162/tacl_a_00605

   in-context retrieval augmentation · experiment

   Why it is here: It evaluates retrieval-conditioned language models and factuality benefits, but does not compare them with fine-tuning alone.

   Abstract · conclusions of the full text read too

   Abstract Retrieval-Augmented Language Modeling (RALM) methods, which condition a language model (LM) on relevant documents from a grounding corpus during generation, were shown to significantly improve language modeling performance. In addition, they can mitigate the problem of factually inaccurate text generation and provide natural source attribution mechanism. Existing RALM approaches focus on modifying the LM architecture in order to facilitate the incorporation of external information, significantly complicating deployment. This paper considers a simple alternative, which we dub In-Context RALM: leaving the LM architecture unchanged and prepending grounding documents to the input, without any further training of the LM. We show that In-Context RALM that builds on off-the-shelf general purpose retrievers provides surprisingly large LM gains across model sizes and diverse corpora. We also demonstrate that the document retrieval and ranking mechanism can be specialized to the RALM setting to further boost performance. We conclude that In-Context RALM has considerable potential to increase the prevalence of LM grounding, particularly in settings where a pretrained LM must be used without modification or even via API access.1

7. [7] Investigating the Factual Knowledge Boundary of Large Language Models with Retrieval Augmentation

   Ruiyang Ren, Yuhao Wang, Yingqi Qu, Wayne Xin Zhao et al. · 2023 · arXiv (Cornell University) preprint · cited 155 times · doi:10.48550/arxiv.2307.11019

   Preprint knowledge-boundary evaluation with RAG · experiment

   Why it is here: It tests how retrieval affects knowledge awareness and handling of internal-external conflicts, relevant to conditions for hallucination reduction.

   Abstract · conclusions of the full text read too

   Large language models (LLMs) have shown impressive prowess in solving a wide range of tasks with world knowledge. However, it remains unclear how well LLMs are able to perceive their factual knowledge boundaries, particularly under retrieval augmentation settings. In this study, we present the first analysis on the factual knowledge boundaries of LLMs and how retrieval augmentation affects LLMs on open-domain question answering (QA), with a bunch of important findings. Specifically, we focus on three research questions and analyze them by examining QA, priori judgement and posteriori judgement capabilities of LLMs. We show evidence that LLMs possess unwavering confidence in their knowledge and cannot handle the conflict between internal and external knowledge well. Furthermore, retrieval augmentation proves to be an effective approach in enhancing LLMs' awareness of knowledge boundaries. We further conduct thorough experiments to examine how different factors affect LLMs and propose a simple method to dynamically utilize supporting documents with our judgement strategy. Additionally, we find that the relevance between the supporting documents and the questions significantly impacts LLMs' QA and judgemental capabilities. The code to reproduce this work is available at https://github.com/RUCAIBox/LLM-Knowledge-Boundary.

8. [8] Making Retrieval-Augmented Language Models Robust to Irrelevant Context

   Ori Yoran, Tomer Wolfson, Ori Ram, Jonathan Berant · 2023 · arXiv (Cornell University) preprint · cited 257 times · doi:10.48550/arxiv.2310.01558

   Preprint irrelevant-context filtering methods · experiment

   Why it is here: It shows that irrelevant retrieved evidence can harm performance and tests mitigations, identifying an important condition for RAG effectiveness.

   Abstract · conclusions of the full text read too

   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.

9. [9] Retrieval-Augmented Generation for Large Language Models: A Survey

   Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia et al. · 2023 · arXiv (Cornell University) preprint · cited 6514 times · doi:10.48550/arxiv.2312.10997

   Review Preprint Survey of RAG architectures · review

   Why it is here: Summarizes retrieval mechanisms and limitations relevant to when RAG may ground outputs, without a direct comparison.

   Abstract · conclusions of the full text read too

   Large Language Models (LLMs) showcase impressive capabilities but encounter challenges like hallucination, outdated knowledge, and non-transparent, untraceable reasoning processes. Retrieval-Augmented Generation (RAG) has emerged as a promising solution by incorporating knowledge from external databases. This enhances the accuracy and credibility of the generation, particularly for knowledge-intensive tasks, and allows for continuous knowledge updates and integration of domain-specific information. RAG synergistically merges LLMs' intrinsic knowledge with the vast, dynamic repositories of external databases. This comprehensive review paper offers a detailed examination of the progression of RAG paradigms, encompassing the Naive RAG, the Advanced RAG, and the Modular RAG. It meticulously scrutinizes the tripartite foundation of RAG frameworks, which includes the retrieval, the generation and the augmentation techniques. The paper highlights the state-of-the-art technologies embedded in each of these critical components, providing a profound understanding of the advancements in RAG systems. Furthermore, this paper introduces up-to-date evaluation framework and benchmark. At the end, this article delineates the challenges currently faced and points out prospective avenues for research and development.

10. [10] Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection

    Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil et al. · 2023 · arXiv (Cornell University) preprint · cited 1091 times · doi:10.48550/arxiv.2310.11511

    Preprint adaptive retrieval and self-reflection · experiment

    Why it is here: It evaluates a retrieval-based approach for factuality, but does not establish how it compares with fine-tuning alone.

    Abstract

    Despite their remarkable capabilities, large language models (LLMs) often produce responses containing factual inaccuracies due to their sole reliance on the parametric knowledge they encapsulate. Retrieval-Augmented Generation (RAG), an ad hoc approach that augments LMs with retrieval of relevant knowledge, decreases such issues. However, indiscriminately retrieving and incorporating a fixed number of retrieved passages, regardless of whether retrieval is necessary, or passages are relevant, diminishes LM versatility or can lead to unhelpful response generation. We introduce a new framework called Self-Reflective Retrieval-Augmented Generation (Self-RAG) that enhances an LM's quality and factuality through retrieval and self-reflection. Our framework trains a single arbitrary LM that adaptively retrieves passages on-demand, and generates and reflects on retrieved passages and its own generations using special tokens, called reflection tokens. Generating reflection tokens makes the LM controllable during the inference phase, enabling it to tailor its behavior to diverse task requirements. Experiments show that Self-RAG (7B and 13B parameters) significantly outperforms state-of-the-art LLMs and retrieval-augmented models on a diverse set of tasks. Specifically, Self-RAG outperforms ChatGPT and retrieval-augmented Llama2-chat on Open-domain QA, reasoning and fact verification tasks, and it shows significant gains in improving factuality and citation accuracy for long-form generations relative to these models.

11. [11] Understanding Retrieval Augmentation for Long-Form Question Answering

    Hung-Ting Chen, Fangyuan Xu, Shane Arora, Eunsol Choi · 2023 · arXiv (Cornell University) preprint · cited 40 times · doi:10.48550/arxiv.2310.12150

    Preprint long-form evidence attribution studies · experiment

    Why it is here: It measures how much generated content is grounded in retrieved documents, relevant to RAG fidelity but not to comparison with fine-tuning.

    Abstract · conclusions of the full text read too

    How retrieved documents are used in language models (LMs) for long-form generation task is understudied. We present two controlled studies on retrieval-augmented LM for long-form question answering (LFQA): one fixing the LM and varying evidence documents and the other fixing evidence documents and varying the LMs. We study various attributes of generated answers (e.g., fluency, length, variance), with an emphasis on the attribution of generated answers to in-context evidence documents. We collect a dataset (SALAD) containing human annotations of sentence-level answer attribution in LFQA and evaluate existing methods for automatically judging attribution. We find that while LMs can leverage relevant in-context documents, the generated answer is only partially attributable towards the documents, especially for LMs trained without retrieval augmentation. Together, our analysis reveals how retrieval augmentation impacts long knowledge-rich text generation and provide directions for future work.

12. [12] A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models

    Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang et al. · 2024 · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · cited 1257 times · doi:10.1145/3637528.3671470

    Recent Review survey of retrieval-augmented LMs · review

    Why it is here: It surveys RAG methods and their role in addressing hallucinations, but the abstract does not specify a direct comparison with fine-tuning.

    Abstract

    As one of the most advanced techniques in AI, Retrieval-Augmented Generation (RAG) can offer reliable and up-to-date external knowledge, providing huge convenience for numerous tasks. Particularly in the era of AI-Generated Content (AIGC), the powerful capacity of retrieval in providing additional knowledge enables RAG to assist existing generative AI in producing high-quality outputs. Recently, Large Language Models (LLMs) have demonstrated revolutionary abilities in language understanding and generation, while still facing inherent limitations such as hallucinations and out-of-date internal knowledge. Given the powerful abilities of RAG in providing the latest and helpful auxiliary information, Retrieval-Augmented Large Language Models (RA-LLMs) have emerged to harness external and authoritative knowledge bases, rather than solely relying on the model's internal knowledge, to augment the quality of the generated content of LLMs. In this survey, we comprehensively review existing research studies in RA-LLMs, covering three primary technical perspectives: Furthermore, to deliver deeper insights, we discuss current limitations and several promising directions for future research. Updated information about this survey can be found at: https://advanced-recommender-systems.github.io/RAG-Meets-LLMs/

13. [13] Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?

    Zorik Gekhman, Gal Yona, Roee Aharoni, Matan Eyal et al. · 2024 · Conference on Empirical Methods in Natural Language Processing (EMNLP) · cited 264 times · doi:10.18653/v1/2024.emnlp-main.444

    Recent Controlled fine-tuning knowledge study · experiment

    Why it is here: Examines how fine-tuning on new facts affects knowledge use and hallucination risk, providing a fine-tuning-only comparison point.

    Abstract

    When large language models are aligned via supervised fine-tuning, they may encounter new factual information that was not acquired through pre-training.It is often conjectured that this can teach the model the behavior of hallucinating factually incorrect responses, as the model is trained to generate facts that are not grounded in its pre-existing knowledge.In this work, we study the impact of such exposure to new knowledge on the capability of the fine-tuned model to utilize its pre-existing knowledge.To this end, we design a controlled setup, focused on closedbook QA, where we vary the proportion of the fine-tuning examples that introduce new knowledge.We demonstrate that large language models struggle to acquire new factual knowledge through fine-tuning, as fine-tuning examples that introduce new knowledge are learned significantly slower than those consistent with the model's knowledge.However, we also find that as the examples with new knowledge are eventually learned, they linearly increase the model's tendency to hallucinate.Taken together, our results highlight the risk in introducing new factual knowledge through fine-tuning, and support the view that large language models mostly acquire factual knowledge through pre-training, whereas finetuning teaches them to use it more efficiently.

14. [14] Domain-Specific Retrieval-Augmented Generation Using Vector Stores, Knowledge Graphs, and Tensor Factorization

    Ryan Barron, Vesselin Grantcharov, Selma Wanna, Maksim E. Eren et al. · 2024 · International Conference on Machine Learning and Applications · cited 39 times · doi:10.1109/icmla61862.2024.00258

    Recent domain-specific structured RAG · experiment

    Why it is here: It evaluates retrieval-based domain adaptation and its effects on QA accuracy, but does not compare against fine-tuning alone.

    Abstract · conclusions of the full text read too

    Large Language Models (LLMs) are pre-trained on large-scale corpora and excel in numerous general natural language processing (NLP) tasks, such as question answering (QA). Despite their advanced language capabilities, when it comes to domain-specific and knowledge-intensive tasks, LLMs suffer from hallucinations, knowledge cut-offs, and lack of knowledge attributions. Additionally, fine tuning LLMs' intrinsic knowledge to highly specific domains is an expensive and time consuming process. The retrieval-augmented generation (RAG) process has recently emerged as a method capable of optimization of LLM responses, by referencing them to a predetermined ontology. It was shown that using a Knowledge Graph (KG) ontology for RAG improves the QA accuracy, by taking into account relevant sub-graphs that preserve the information in a structured manner. In this paper, we introduce SMART-SLIC, a highly domain-specific LLM framework, that integrates RAG with KG and a vector store (VS) that store factual domain specific information. Importantly, to avoid hallucinations in the KG, we build these highly domain-specific KGs and VSs without the use of LLMs, but via NLP, data mining, and nonnegative tensor factorization with automatic model selection. Pairing our RAG with a domain-specific: (i) KG (containing structured information), and (ii) VS (containing unstructured information) enables the development of domain-specific chat-bots that attribute the source of information, mitigate hallucinations, lessen the need for fine-tuning, and excel in highly domain-specific question answering tasks. We pair SMART-SLIC with chain-of-thought prompting agents. The framework is designed to be generalizable to adapt to any specific or specialized domain. In this paper, we demonstrate the question answering capabilities of our framework on a corpus of scientific publications on malware analysis and anomaly detection.

15. [15] Enhancing LLM Factual Accuracy with RAG to Counter Hallucinations: A Case Study on Domain-Specific Queries in Private Knowledge-Bases

    Jiarui Li, Ye Yuan, Zehua Zhang · 2024 · arXiv (Cornell University) preprint · cited 176 times · doi:10.48550/arxiv.2403.10446

    Recent Preprint Private-knowledge-base RAG case study · experiment

    Why it is here: Evaluates RAG for domain-specific, time-sensitive facts and notes fine-tuning limitations, but provides no clear direct comparison.

    Abstract · conclusions of the full text read too

    We proposed an end-to-end system design towards utilizing Retrieval Augmented Generation (RAG) to improve the factual accuracy of Large Language Models (LLMs) for domain-specific and time-sensitive queries related to private knowledge-bases. Our system integrates RAG pipeline with upstream datasets processing and downstream performance evaluation. Addressing the challenge of LLM hallucinations, we finetune models with a curated dataset which originates from CMU's extensive resources and annotated with the teacher model. Our experiments demonstrate the system's effectiveness in generating more accurate answers to domain-specific and time-sensitive inquiries. The results also revealed the limitations of fine-tuning LLMs with small-scale and skewed datasets. This research highlights the potential of RAG systems in augmenting LLMs with external datasets for improved performance in knowledge-intensive tasks. Our code and models are available on Github.

16. [16] Fine Tuning vs. Retrieval Augmented Generation for Less Popular Knowledge

    Heydar Soudani, Evangelos Kanoulas, Faegheh Hasibi · 2024 · Annual International ACM SIGIR Conference on Research and Development in Information Retrieval in the Asia Pacific Region (SIGIR-AP) · cited 140 times · doi:10.1145/3673791.3698415

    Recent Compare retrieval and fine-tuning · experiment

    Why it is here: Directly evaluates RAG against fine-tuning for low-frequency knowledge and reports when retrieval performs better.

    Abstract · conclusions of the full text read too

    Language Models (LMs) memorize a vast amount of factual knowledge, exhibiting strong performance across diverse tasks and domains. However, it has been observed that the performance diminishes when dealing with less-popular or low-frequency concepts and entities, for example in domain specific applications. The two prominent approaches to enhance the performance of LMs on low-frequent topics are: Retrieval Augmented Generation (RAG) and fine-tuning (FT) over synthetic data. This paper explores and evaluates the impact of RAG and FT on customizing LMs in handling low-frequency entities on question answering tasks. We conduct extensive experiments on twelve LMs of varying size and type and different FT methods, data augmentation, and retrieval models. Our findings indicate that while FT boosts the performance across entities of varying popularity, RAG surpasses FT by a large margin particularly for least popular factual knowledge. Additionally, the success of both RAG and FT approaches is amplified by improving retrieval and data augmentation techniques. Fine tuning, while beneficial for small LMs, requires extensive resources. To address this issue, we propose the new Stimulus RAG approach that surpasses the effectiveness of fine tuning based approaches, thereby eliminating the need for the costly data augmentation and fine tuning step for enriching LMs with less popular factual knowledge.

17. [17] Fine-Tuning or Fine-Failing? Debunking Performance Myths in Large Language Models

    Scott Barnett, Zach Brannelly, Stefanus Kurniawan, Sheng Wong · 2024 · arXiv (Cornell University) preprint · cited 22 times · doi:10.48550/arxiv.2406.11201

    Recent Preprint Fine-tuning within RAG pipelines · experiment

    Why it is here: Studies fine-tuning in RAG settings, but the abstract does not establish a direct comparison with fine-tuning alone.

    Abstract · conclusions of the full text read too

    Large Language Models (LLMs) have the unique capability to understand and generate human-like text from input queries. When fine-tuned, these models show enhanced performance on domain-specific queries. OpenAI highlights the process of fine-tuning, stating: "To fine-tune a model, you are required to provide at least 10 examples. We typically see clear improvements from fine-tuning on 50 to 100 training examples, but the right number varies greatly based on the exact use case." This study extends this concept to the integration of LLMs within Retrieval-Augmented Generation (RAG) pipelines, which aim to improve accuracy and relevance by leveraging external corpus data for information retrieval. However, RAG's promise of delivering optimal responses often falls short in complex query scenarios. This study aims to specifically examine the effects of fine-tuning LLMs on their ability to extract and integrate contextual data to enhance the performance of RAG systems across multiple domains. We evaluate the impact of fine-tuning on the LLMs' capacity for data extraction and contextual understanding by comparing the accuracy and completeness of fine-tuned models against baseline performances across datasets from multiple domains. Our findings indicate that fine-tuning resulted in a decline in performance compared to the baseline models, contrary to the improvements observed in standalone LLM applications as suggested by OpenAI. This study highlights the need for vigorous investigation and validation of fine-tuned models for domain-specific tasks.

18. [18] Fine-Tuning or Retrieval? Comparing Knowledge Injection in LLMs

    Oded Ovadia, Menachem Brief, Moshik Mishaeli, Oren Elisha · 2024 · Conference on Empirical Methods in Natural Language Processing (EMNLP) · cited 340 times · doi:10.18653/v1/2024.emnlp-main.15

    Recent Compare fine-tuning and retrieval · experiment

    Why it is here: Directly compares knowledge injection by fine-tuning and RAG across tasks and reports RAG advantages.

    Abstract

    Large language models (LLMs) encapsulate a vast amount of factual information within their pre-trained weights, as evidenced by their ability to answer diverse questions across different domains.However, this knowledge is inherently limited, relying heavily on the characteristics of the training data.Consequently, using external datasets to incorporate new information or refine the capabilities of LLMs on previously seen information poses a significant challenge.In this study, we compare two common approaches: unsupervised fine-tuning and retrieval-augmented generation (RAG).We evaluate both approaches on a variety of knowledge-intensive tasks across different topics.Our findings reveal that while unsupervised fine-tuning offers some improvement, RAG consistently outperforms it, both for existing knowledge encountered during training and entirely new knowledge.Moreover, we find that LLMs struggle to learn new factual information through unsupervised fine-tuning, and that exposing them to numerous variations of the same fact during training could alleviate this problem.

19. [19] Grounding by Trying: LLMs with Reinforcement Learning-Enhanced Retrieval

    Sheryl Hsu, Omar Khattab, Chelsea Finn, Archit Sharma · 2024 · arXiv (Cornell University) preprint · cited 24 times · doi:10.48550/arxiv.2410.23214

    Recent Preprint reinforcement-learned retrieval queries · experiment

    Why it is here: It evaluates retrieval improvements and downstream generation, offering evidence about retrieval-based mitigation but not a fine-tuning-only comparison.

    Abstract · conclusions of the full text read too

    The hallucinations of large language models (LLMs) are increasingly mitigated by allowing LLMs to search for information and to ground their answers in real sources. Unfortunately, LLMs often struggle with posing the right search queries, especially when dealing with complex or otherwise indirect topics. Observing that LLMs can learn to search for relevant facts by $\textit{trying}$ different queries and learning to up-weight queries that successfully produce relevant results, we introduce $\underline{Le}$arning to $\underline{Re}$trieve by $\underline{T}$rying (LeReT), a reinforcement learning framework that explores search queries and uses preference-based optimization to improve their quality. LeReT can improve the absolute retrieval accuracy by up to 29% and the downstream generator evaluations by 17%. The simplicity and flexibility of LeReT allows it to be applied to arbitrary off-the-shelf retrievers and makes it a promising technique for improving general LLM pipelines. Project website: http://sherylhsu.com/LeReT/.

20. [20] Lynx: An Open Source Hallucination Evaluation Model

    Ravi, Selvan Sunitha, Mielczarek, Bartosz, A. Kannappan, Douwe Kiela et al. · 2024 · arXiv (Cornell University) preprint · cited 82 times · doi:10.48550/arxiv.2407.08488

    Recent Preprint RAG hallucination evaluation benchmark · method

    Why it is here: Introduces HaluBench and an evaluator for unsupported or conflicting RAG answers, useful for measuring outcomes.

    Abstract · conclusions of the full text read too

    Retrieval Augmented Generation (RAG) techniques aim to mitigate hallucinations in Large Language Models (LLMs). However, LLMs can still produce information that is unsupported or contradictory to the retrieved contexts. We introduce LYNX, a SOTA hallucination detection LLM that is capable of advanced reasoning on challenging real-world hallucination scenarios. To evaluate LYNX, we present HaluBench, a comprehensive hallucination evaluation benchmark, consisting of 15k samples sourced from various real-world domains. Our experiment results show that LYNX outperforms GPT-4o, Claude-3-Sonnet, and closed and open-source LLM-as-a-judge models on HaluBench. We release LYNX, HaluBench and our evaluation code for public access.

21. [21] A Retrieval-augmented Generation Framework with Retriever and Generator Modules for Enhancing Factual Consistency

    Yang Zhang · 2025 · Applied and Computational Engineering · cited 2 times · doi:10.54254/2755-2721/2025.tj24496

    Recent Retriever-generator factual grounding · experiment

    Why it is here: Reports reduced hallucination with RAG on QA benchmarks, but lacks a fine-tuning-only comparison.

    Abstract · conclusions of the full text read too

    Large Language Models (LLMs) are powerful but often produce factually incorrect content (hallucinations), limiting their reliability in knowledge-intensive tasks. Retrieval-augmented generation (RAG) is a promising approach to mitigate this issue by grounding LLM outputs in external knowledge sources. The paper proposes an RAG framework integrating a retriever and generator module to improve factual consistency. The retriever first identifies relevant documents from large-scale datasets, and the generator then produces context-aware responses based on the retrieved evidence. This study evaluates the approach on open-domain question answering benchmarks, including Natural Questions, Microsoft Machine Reading Comprehension Dataset (MS MARCO), and Covid-19 Open Research Dataset (CORD-19). The RAG-augmented model significantly reduces the hallucination rate from 68% to 10% and increases the Knowledge F1 score from 17.7 to 26.0, outperforming a baseline LLM without retrieval. These results demonstrate that augmenting LLMs with retrieval substantially enhances their factual accuracy and reliability. This improvement is significant for time-sensitive and domain-specific applications, such as healthcare and legal contexts, where up-to-date and accurate information is critical.

22. [22] Hybrid Retrieval for Hallucination Mitigation in Large Language Models: A Comparative Analysis

    Chandana Sree Mala, Gizem Gezici, Fosca Giannotti · 2025 · arXiv · arXiv: 2504.05324v1

    Recent Preprint Compare hybrid and baseline retrieval · experiment

    Why it is here: Tests retrieval designs for reducing hallucinations, but does not compare RAG with fine-tuning.

    Abstract · conclusions of the full text read too

    Large Language Models (LLMs) excel in language comprehension and generation but are prone to hallucinations, producing factually incorrect or unsupported outputs. Retrieval Augmented Generation (RAG) systems address this issue by grounding LLM responses with external knowledge. This study evaluates the relationship between retriever effectiveness and hallucination reduction in LLMs using three retrieval approaches: sparse retrieval based on BM25 keyword search, dense retrieval using semantic search with Sentence Transformers, and a proposed hybrid retrieval module. The hybrid module incorporates query expansion and combines the results of sparse and dense retrievers through a dynamically weighted Reciprocal Rank Fusion score. Using the HaluBench dataset, a benchmark for hallucinations in question answering tasks, we assess retrieval performance with metrics such as mean average precision and normalised discounted cumulative gain, focusing on the relevance of the top three retrieved documents. Results show that the hybrid retriever achieves better relevance scores, outperforming both sparse and dense retrievers. Further evaluation of LLM-generated answers against ground truth using metrics such as accuracy, hallucination rate, and rejection rate reveals that the hybrid retriever achieves the highest accuracy on fails, the lowest hallucination rate, and the lowest rejection rate. These findings highlight the hybrid retriever's ability to enhance retrieval relevance, reduce hallucination rates, and improve LLM reliability, emphasising the importance of advanced retrieval techniques in mitigating hallucinations and improving response accuracy.

23. [23] Probabilistic distances-based hallucination detection in LLMs with RAG

    Rodion Oblovatny, Kuleshova, Alexandra, Polev, Konstantin, Zaytsev, Alexey · 2025 · arXiv (Cornell University) preprint · cited 5 times · doi:10.48550/arxiv.2506.09886

    Recent Preprint Hidden-state distance detection · experiment

    Why it is here: Studies hallucination detection specifically in RAG, contributing evidence about evaluation rather than comparative mitigation.

    Abstract · conclusions of the full text read too

    Detecting hallucinations in large language models (LLMs) is critical for their safety in many applications. Without proper detection, these systems often provide harmful, unreliable answers. In recent years, LLMs have been actively used in retrieval-augmented generation (RAG) settings. However, hallucinations remain even in this setting, and while numerous hallucination detection methods have been proposed, most approaches are not specifically designed for RAG systems. To overcome this limitation, we introduce a hallucination detection method based on estimating the distances between the distributions of prompt token embeddings and language model response token embeddings. The method examines the geometric structure of token hidden states to reliably extract a signal of factuality in text, while remaining friendly to long sequences. Extensive experiments demonstrate that our method achieves state-of-the-art or competitive performance. It also has transferability from solving the NLI task to the hallucination detection task, making it a fully unsupervised and efficient method with a competitive performance on the final task.

24. [24] Augmenting Large Language Models with External Data Sources: A Systematic Review of Methodologies, Performance Metrics, and Information Fidelity

    Soham Mukherjee, John D. Le, Chau Nguyen, Thai T. Vu · 2026 · Knowledge · cited 0 times · doi:10.3390/knowledge6030013

    Recent Review systematic review of RAG and fine-tuning · review

    Why it is here: It directly compares the two dominant knowledge-augmentation paradigms through information fidelity, the core of the research question.

    Abstract · conclusions of the full text read too

    Large Language Models (LLMs) have emerged as transformative tools across various domains, exhibiting remarkable capabilities in natural language processing and generation. However, their reliance on static pre-training data limits their ability to access up-to-date and domain-specific information. The existing research often treats augmentation strategies in isolation, and limited efforts have been made to systematically compare them through the lens of information integrity. This review focuses specifically on Retrieval-Augmented Generation (RAG) and fine-tuning, identifying them as the two dominant paradigms for integrating external knowledge: RAG for retrieval-based context injection and fine-tuning for parametric knowledge adaptation. While existing surveys predominantly focus on performance metrics like accuracy or latency, this paper addresses the critical gap of data fidelity—the preservation of truthfulness, integrity, and fairness during augmentation. We systematically synthesize empirical findings from diverse methodologies to determine how each approach mitigates hallucinations and bias. By comparing the trade-offs between retrieval-based context injection and parametric knowledge adaptation, this survey provides unique value to readers by providing a structured taxonomy, a unified evaluation framework, and actionable insights to guide future research and practical deployment of robust, high-fidelity LLMs.

25. [25] Detecting and Repairing Hallucinations in Retrieval-Augmented Generation

    Sai Krishna Reddy Mulakkayala, Niki van Stein, Aske Plaat · 2026 · arXiv · arXiv: 2608.29307v1

    Recent Preprint Claim-level RAG answer repair · experiment

    Why it is here: Evaluates hallucination repair against retrieved sources, informing RAG reliability but not comparison with fine-tuning.

    Abstract · conclusions of the full text read too

    Language models increasingly answer questions by consulting retrieved documents rather than memory alone, a design now common in search assistants and enterprise knowledge tools. Grounding a model in retrieved text reduces unsupported statements but does not eliminate them, and a reader cannot tell a grounded sentence from an invented one. Most research on this problem stops at detection, yet flagging a faulty answer changes nothing for the person reading it, and little is known about which action should follow. Using RAGTruth, a benchmark whose unsupported passages are annotated by hand, we split each flagged answer into individual factual claims, check each against the retrieved source, and compare leaving the answer untouched with three repair strategies of increasing richness: deleting an unsupported claim, replacing it with source text, and rewriting it. Three language models from different families judge the 916 repaired answers. Every strategy reduces the proportion of answers judged to contain unsupported content, and all three judges agree on the ordering. Deletion achieves the largest reduction while retaining least of the original answer, at 64.3% of the text, whereas rewriting retains 80.1% and reduces least. Repair is not confined to faulty answers: 83.5% of answers annotated clean are edited too. The strategies occupy different points on a grounding preservation trade-off rather than forming a quality ranking, and choosing between them needs evidence about answer usefulness that automatic metrics cannot supply.

26. [26] Hallucination in Large Language Models and Retrieval-Augmented Generation: Mechanisms, Mitigation, and Evaluation

    Haopeng Yang · 2026 · Theoretical and Natural Science · cited 0 times · doi:10.54254/2753-8818/2026.dl34055

    Recent Review Review RAG hallucination mechanisms · review

    Why it is here: Explains when retrieval can fail to prevent hallucinations, but offers no direct RAG-versus-fine-tuning comparison.

    Abstract · conclusions of the full text read too

    Large language models have demonstrated strong generative capability in question answering, dialogue, and other knowledge-intensive tasks. However, their outputs remain vulnerable to hallucination, including factual errors, unsupported claims, spurious citations, and distorted reasoning. Retrieval-augmented generation (RAG) has been proposed as a practical remedy because it supplements parametric knowledge with external evidence retrieved at inference time. Yet RAG does not guarantee truthfulness or attribution by default. Errors may arise during query formulation, document retrieval, evidence aggregation, and answer grounding. This paper reviews the relationship between hallucination and RAG from three perspectives: mechanism, mitigation, and evaluation. First, the major causes of hallucination in both vanilla LLMs and RAG-enhanced systems are analyzed. Second, the principal mitigation strategies are organized into retrieval optimization, evidence-grounded generation, and post-generation verification. Third, the main evaluation dimensions are examined, including factuality, faithfulness, attribution quality, and retrieval relevance. It is argued that RAG should not be treated as a complete solution to hallucination. Its value lies in enabling externally grounded generation, but its effectiveness depends on the reliability of retrieval, the fidelity of evidence use, and the rigor of evaluation. Future work should prioritize attribution-aware generation, conflict-sensitive reasoning, and unified evaluation protocols for trustworthy LLM systems.

27. [27] Integrating Fine-Tuning and Retrieval-Augmented Generation for Healthcare AI Systems: A Scoping Review

    Bernardo Gabriele Collaço, Prabha Srinivasagam, Cesar Abraham Gomez-Cabello, Syed Ali Haider et al. · 2026 · Bioengineering · cited 8 times · doi:10.3390/bioengineering13020225

    Recent Review healthcare hybrid-methods scoping review · review

    Why it is here: It synthesizes empirical healthcare evidence on combining fine-tuning and RAG, relevant to their relative benefits and application conditions.

    Abstract · conclusions of the full text read too

    (1) Background: Large language models (LLMs) show promise in healthcare but are constrained by hallucinations, static knowledge, and limited domain specificity. Fine-tuning (FT) and retrieval-augmented generation (RAG) offer complementary solutions, with FT embedding domain reasoning and RAG enabling dynamic, up-to-date knowledge access. Hybrid FT + RAG frameworks have been proposed to improve factual accuracy and clinical reliability. This scoping review synthesizes current evidence on such hybrids in healthcare AI. (2) Methods: The search across PubMed, IEEE Xplore, Google Scholar, and Embase identified studies implementing explicit FT + RAG hybrids in healthcare or biomedical tasks. Eligible studies reported empirical evaluations of LLM performance or behavior. Data were extracted on base models, FT strategies, RAG architectures, applications, and performance outcomes. (3) Results: Seven studies met inclusion criteria. FT + RAG systems consistently outperformed FT-only or RAG-only approaches across QA, clinical summarization, report generation, and decision support tasks. Parameter-efficient FT methods (e.g., LoRA) were common, while RAG implementations varied (dense, hybrid, hierarchical, multimodal, federated). Reported benefits included improved accuracy, reduced hallucination, and greater clinician preference and feasibility in protected settings. (4) Conclusions: FT + RAG frameworks represent a promising direction for clinically grounded healthcare AI, combining domain-specific reasoning with transparent, up-to-date retrieval. Future work should prioritize standardized evaluation, workflow integration, and governance to enable safe deployment.

28. [28] LoRA-Based Fine-Tuning of Local LLMs for Hallucination Detection in Indonesian RAG Systems

    I Ketut Resika Arthana, Nyoman Gunantara, Made Sudarma, Made Sukarsa · 2026 · International Journal of Advanced Computer Science and Applications · cited 1 time · doi:10.14569/ijacsa.2026.0170389

    Recent LoRA-based RAG hallucination detection · experiment

    Why it is here: Evaluates hallucination detection in Indonesian RAG systems using fine-tuning, but does not compare fine-tuning as an alternative to RAG.

    Abstract · conclusions of the full text read too

    Retrieval Augmented Generation (RAG) improves the factual grounding of Large Language Models (LLMs) by incorporating external knowledge. However, RAG systems may still generate hallucinated responses, and this issue remains underexplored in Indonesian language settings, particularly in settings where local deployment is preferred. This study proposes a hallucination detection approach for Indonesian RAG systems using Low Rank Adaptation (LoRA) fine-tuning. To support this objective, the study constructs a dataset in the Human-Computer Interaction domain consisting of 908 context, question, and answer pairs. The dataset is classified into four categories: FACT-H, FAITH-H, LOG-H, and FAITHFUL. Three local LLMs, namely, Gemma-7B-it, LlaMA-2-7B chat, and Phi-3-medium-4k-instruct, were evaluated using 5-fold cross-validation. The results show that Gemma-7B-it achieved the best performance in the four-class setting, with a Macro F1 score of 0.846. In the binary classification setting, Gemma achieved an accuracy of 98.1 per cent. Further analysis shows that Gemma was particularly effective in recognizing FAITHFUL, FAITH-H, and FACT-H, while LOG-H remained the most difficult class to distinguish consistently.

29. [29] Retrieval-Augmented Generation (RAG) vs. Fine-Tuning: A Systematic Review of Hallucination Mitigation in Medical LLMs

    Hassan Jubair · 2026 · cited 1 time · doi:10.22541/au.176903295.54191070/v1

    Recent Review Preprint Systematic medical strategy comparison · review

    Why it is here: Directly synthesizes evidence comparing RAG and fine-tuning for hallucination mitigation in medical applications.

    Abstract

    Background: The integration of Large Language Models (LLMs) into healthcare is currently hindered by the critical risk of "hallucinations", the generation of plausible but factually incorrect medical information.Two primary adaptation strategies have emerged to address this: Fine-Tuning (FT), which updates model parameters to internalize domain knowledge, and Retrieval-Augmented Generation (RAG), which grounds model outputs in external evidence at inference time.Objective: This systematic review synthesizes findings from 50 peer-reviewed studies published between 2023 and 2025 to evaluate the comparative effectiveness of RAG versus Fine-Tuning in medical contexts.Results: Evidence consistently demonstrates that RAG is the superior strategy for clinical safety, significantly reducing hallucination rates by dynamically accessing up-to-date knowledge without the need for retraining.Conversely, while Fine-Tuning excels at adapting to domain-specific syntax and clinical reasoning styles, it incurs high computational costs and carries a significant risk of "catastrophic forgetting," where general knowledge is degraded.Conclusion: We conclude that while Fine-Tuning remains valuable for specialization, RAG provides the necessary factual guardrails for patient-facing applications.Future development should prioritize "Hybrid" frameworks that synergize the reasoning depth of Fine-Tuning with the factual integrity of RAG.

30. [30] Retromorphic Testing with Hierarchical Verification for Hallucination Detection in RAG

    Boxi Yu, Yuzhong Zhang, Liting Lin, Lionel Briand et al. · 2026 · arXiv (Cornell University) preprint · cited 3 times · doi:10.48550/arxiv.2603.27752

    Recent Preprint Claim-level context verification method · method

    Why it is here: Provides a way to detect unsupported or contradicted claims in RAG outputs, relevant to measuring grounding.

    Abstract · conclusions of the full text read too

    Large language models can still hallucinate in retrieval-augmented generation (RAG), producing claims that are unsupported by or conflict with the retrieved context. Detecting such errors remains challenging when faithfulness is judged solely against the retrieved context: many existing detectors return holistic answer-level scores, while others target open-domain factuality or fail to provide evidence-grounded diagnostics. We present RT4CHART, a retromorphic testing framework for context-faithfulness assessment. RT4CHART decomposes an answer into independently verifiable claims, performs hierarchical local-to-global verification against the retrieved context, and assigns each claim one of three labels: entailed, contradicted, or baseless. It further maps these claim-level decisions back to specific answer spans and returns explicit context-side evidence, enabling fine-grained auditing rather than opaque scoring. We evaluate RT4CHART on RAGTruth++ (408 samples) and our re-annotated RAGTruth-Enhance (2,675 samples). RT4CHART achieves the best answer-level hallucination-detection F1 score among the evaluated baselines. On RAGTruth++, it attains a precision of 0.845, a recall of 0.718, and an F1 score of 0.776, representing an 83% relative improvement over the strongest baseline. It also achieves a span-level F1 score of 47.5% on RAGTruth-Enhance. Ablation studies show that claim-based local processing drives most of the observed improvement, while global verification provides selective benefits across datasets. Finally, our re-annotation identifies 1.68X more hallucination cases than the original labels, suggesting that commonly used benchmarks substantially underestimate the prevalence of hallucination.