Revue de littérature partagée · Faire la revue de littérature de votre propre question
RAG et hallucinations des grands modèles de langage
Does retrieval-augmented generation reduce the hallucinations of large language models compared with fine-tuning alone?
Comment cette revue a été construite
305 articles trouvés, les 80 plus proches de la question triés sur leur résumé, 63 jugés pertinents, 30 retenus. Publiés de 2021 à 2026. 1 articles fondateurs, 23 des trois dernières années, 9 revues de littérature.
Publiés dans une revue ou des actes : 22 · préprints : 8 · dépôts que personne n’a relus : 0. Les conclusions de 12 textes intégraux en accès libre ont aussi été lues.
19 des 19 résultats sont appuyés par une phrase trouvée mot pour mot dans un article. De même que 5 des 6 points d’accord, de désaccord et de lacunes.
Une synthèse exploratoire, pas une revue systématique : les articles sont lus à partir de leurs résumés et des conclusions de leurs textes intégraux en accès libre, et des articles qu’OpenAlex et arXiv n’indexent pas, ou que nos recherches formulent autrement, peuvent manquer.
Recherches et sources
Inclusion : Include empirical comparisons of retrieval-augmented generation and fine-tuning, along with studies of grounding mechanisms, hallucination measures, benchmarks, and task or domain settings in large language models.
Concepts clés : Retrieval-augmented generation, Fine-tuning, Hallucination, Large language models
Requêtes :
- RAG hallucination large language models
- retrieval grounding factuality unsupported claims
- fine-tuning RAG hallucination comparison
- retrieval augmentation factual consistency benchmarks
- RAG hallucination knowledge-intensive question answering
- retrieval generation error mechanisms evaluation
- OpenAlex, par pertinence : 141
- OpenAlex, trois dernières années : 85
- arXiv : 69
- OpenAlex, revues de littérature : 15
- OpenAlex, les plus cités : 25
- Cités par plusieurs des articles les plus proches : 50
- Articles récents qui citent les fondateurs : 25
Un article trouvé par plusieurs recherches est compté dans chacune.
En bref
Les résultats disponibles ne permettent pas d’affirmer que le RAG réduit systématiquement davantage les hallucinations que le fine-tuning seul : les comparaisons directes sont peu nombreuses et portent sur des formes différentes d’adaptation et d’évaluation. [12][13][17][24] Une expérience publiée favorise le RAG face au fine-tuning non supervisé pour intégrer des connaissances, mais une autre étude trouve de meilleures performances avec le modèle ajusté ; ces résultats concernent surtout la réussite aux tâches, qui ne mesure pas exclusivement les hallucinations. [12][17] En biomédecine, une méta-analyse rapporte un bénéfice global du RAG face aux modèles de base, sans établir sa supériorité sur le fine-tuning seul, tandis qu’une revue de sept études favorise les systèmes hybrides FT + RAG. [20][24] Le bénéfice dépend fortement de la pertinence des sources, de leur utilisation fidèle et du protocole d’évaluation ; les méthodes adaptatives, la vérification et l’apprentissage spécifique à l’ancrage documentaire cherchent précisément à corriger les défaillances du RAG simple. [6][8][15][23][30]
Ce résumé repose sur les résultats ci-dessous, chacun avec les phrases qui l’appuient.
Approches comparées
| Approche | Articles | Points forts | Limites |
|---|---|---|---|
| Comparer l’injection de connaissances par RAG, fine-tuning et combinaison des deux | [11][12][13][17][24][28] | Ce groupe distingue utilement l’acquisition de faits, l’adaptation du comportement et leur combinaison. | Les variations de modèles, de fine-tuning et de critères empêchent d’en déduire un effet comparatif unique sur les hallucinations. |
| Ancrer les réponses dans des sources adaptées à la tâche | [1][2][9][14][20][27] | L’accès externe est particulièrement utile pour des connaissances spécialisées, structurées ou susceptibles d’être actualisées. | Les bénéfices observés ne permettent généralement ni d’isoler le rôle du RAG dans les pipelines combinés ni de conclure à sa supériorité sur le fine-tuning. |
| Adapter la recherche et examiner les preuves plutôt que récupérer systématiquement | [3][6][10][25][29] | Ces approches cherchent à éviter l’incorporation aveugle de passages inutiles et à améliorer la sélection des preuves. | Elles sont surtout comparées à d’autres variantes de RAG, et leurs gains ne tranchent donc pas la comparaison générale avec le fine-tuning seul. |
| Entraîner le générateur à reconnaître et à éviter les hallucinations dans un contexte RAG | [15][18][26] | L’apprentissage vise directement la fidélité aux preuves et la résistance aux contextes trompeurs. | Ces expériences établissent l’intérêt d’ajuster un système RAG plutôt qu’une supériorité autonome du RAG sur le fine-tuning. |
| Évaluer la fidélité documentaire, la robustesse et les effets de présentation | [4][7][8][16][27][30] | Ces protocoles identifient des défaillances que l’exactitude agrégée seule peut masquer. | Ils ne produisent pas une estimation harmonisée de l’effet du RAG face au fine-tuning, et aucune conclusion substantielle ne peut être tirée du seul extrait fourni pour le papier 16. |
| Comprendre les hallucinations comme un problème de pipeline complet | [5][19][21][22][23] | Cette lecture explique pourquoi les résultats varient entre systèmes portant tous l’étiquette RAG. | Ces synthèses ne fournissent pas d’effet comparatif homogène permettant de classer universellement RAG, fine-tuning et systèmes hybrides. |
Comparer l’injection de connaissances par RAG, fine-tuning et combinaison des deux
Ces travaux sont les plus proches de la comparaison demandée, mais ils ne comparent ni les mêmes formes de fine-tuning ni les mêmes tâches. L’étude publiée sur le fine-tuning non supervisé favorise le RAG, alors que l’expérience de 2025 favorise le modèle ajusté ; les travaux sur les hybrides suggèrent plutôt une complémentarité, avec un appui encore limité dans certains cas à des prépublications non évaluées.
- Dans une comparaison expérimentale publiée, le RAG surpasse systématiquement le fine-tuning non supervisé sur les tâches de connaissances étudiées, ce qui ne constitue pas une comparaison avec toutes les formes de fine-tuning. [12]
« 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. » [12]
- Une autre expérience publiée rapporte de meilleurs résultats pour le modèle ajusté et ne constate pas d’amélioration significative par les solutions RAG étudiées, sans permettre de généraliser ce classement à tous les systèmes. [17]
« Our findings show that while fine-tuned LLMs outperform baseline models, incorporating RAG solutions did not result in significant performance improvements, suggesting that the incorporation of external knowledge may not always align with the needs of the task. » [17]
- Les approches hybrides obtiennent les meilleurs résultats dans une prépublication sur CRAG et dans les sept études retenues par une revue exploratoire publiée en santé. [13][24]
« Our results show that the hybrid approach using both RAG and fine-tuning performs best in CRAG. » [13] · texte intégral
1 autre phrase à l’appui
« FT + RAG systems consistently outperformed FT-only or RAG-only approaches across QA, clinical summarization, report generation, and decision support tasks. » [24]
- Deux prépublications non évaluées apportent des arguments supplémentaires en faveur du RAG, mais l’une souligne les limites de l’ajustement génératif sur de petits jeux biaisés et l’autre avance une conclusion de sécurité clinique plus forte que ce qu’établissent les comparaisons hétérogènes du corpus. [11][28]
« The ablation studies also indicate the necessity of finetuning the embedding model, and the limitations of finetuning the large generative model with small and biased datasets. » [11] · texte intégral
1 autre phrase à l’appui
« 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. » [28]
Ancrer les réponses dans des sources adaptées à la tâche
Ces travaux testent l’accès à des connaissances externes en dialogue, finance, médecine, description visuelle et raisonnement multi-étapes. Ils documentent des gains parfois importants, mais généralement face à un modèle de base ou à une stratégie de prompting plutôt que face au fine-tuning seul.
- En dialogue, une étude publiée constate une réduction substantielle des hallucinations vérifiée par évaluation humaine, tandis qu’une prépublication financière rapporte l’efficacité du RAG et des appels d’outils pour fournir des informations correctes et actualisées. [1][2]
« 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. » [1]
1 autre phrase à l’appui
« To mitigate hallucinations, we show the effectiveness of the RAG method and prompt-based tool learning to generate correct function calls, thereby ensuring the provision of factually correct and up-to-date information in the finance domain. » [2] · texte intégral
- En médecine, MEDRAG rapporte jusqu’à 18 % d’amélioration d’exactitude face au prompting par chaîne de pensée, et la méta-analyse de vingt études trouve un odds ratio global de 1,35, deux résultats qui ne quantifient pas directement une réduction des hallucinations face au fine-tuning. [9][20]
« Overall, MEDRAG improves the accuracy of six different LLMs by up to 18% over chain-of-thought prompting, elevating the performance of GPT-3.5 and Mixtral to GPT-4level. » [9]
1 autre phrase à l’appui
« The pooled effect size was 1.35, with a 95% confidence interval of 1.19-1.53, indicating a statistically significant effect (P = .001). » [20]
- Les extensions à la description spatiale visuelle et aux connaissances structurées Wikidata rapportent également des bénéfices, mais chacune repose ici sur une seule étude et l’expérience visuelle combine plusieurs interventions. [14][27]
« In this paper, we apply Retrieval-Augmented Generation (RAG) technology in guiding Multimodal Large Language Models (MLLMs) for the task of VSD, complemented by an Adaptive Hallucination Corrector, and further fine-tuning them to bolster semantic understanding and overall model efficacy. » [14]
1 autre phrase à l’appui
« RAG consistently surpasses the baseline across all four evaluators. » [27]
Adapter la recherche et examiner les preuves plutôt que récupérer systématiquement
Ces méthodes ajoutent une boucle de recherche-génération, une réflexion sur les passages, des notes de lecture ou un vote entre plusieurs réponses. Elles traitent la pertinence documentaire comme une décision à contrôler, plutôt que comme une propriété acquise du RAG.
- ITER-RETGEN utilise les réponses intermédiaires pour améliorer la recherche et obtient des performances supérieures ou comparables aux références augmentées sur plusieurs tâches, sans fournir ici de comparaison directe avec le fine-tuning seul. [3]
« We evaluate ITER-RETGEN on multi-hop question answering, fact verification, and commonsense reasoning, and show that it can flexibly leverage parametric knowledge and non-parametric knowledge, and is superior to or competitive with state-of-the-art retrieval-augmented baselines while causing fewer overheads of retrieval and generation. » [3]
- Self-RAG, présenté dans une prépublication non évaluée, apprend à récupérer et à réfléchir à la demande, tandis que Chain-of-Note améliore des modèles augmentés déjà ajustés sur quatre benchmarks de questions-réponses. [6][10]
« 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. » [6]
1 autre phrase à l’appui
« Our experiments across four open-domain QA benchmarks show that fine-tuned RALMs equipped with CON significantly outperform standard fine-tuned RALMs. » [10]
- Une étude publiée propose un vote parallèle sans entraînement et rapporte une exactitude comparable à des systèmes plus complexes, tandis qu’une revue distingue les fonctions de réflexion de Self-RAG et de correction documentaire de Corrective RAG. [25][29]
« Our experiments across six datasets demonstrate that VOTE-RAG achieves robust factual accuracy comparable to more complex frameworks while remaining architecturally simpler and fully parallelizable. » [25] · texte intégral
1 autre phrase à l’appui
« Corrective RAG pays more attention to the quality of retrieved information; it checks and corrects poor retrieved content before the generation stage. » [29]
Entraîner le générateur à reconnaître et à éviter les hallucinations dans un contexte RAG
Ces travaux font du fine-tuning un complément du RAG en apprenant au modèle à mieux exploiter des contextes imparfaits ou à détecter ses productions non étayées. Ils déplacent la question de « RAG ou fine-tuning » vers celle de l’objectif d’apprentissage qui rend le RAG plus fiable.
- RAG-HAT utilise des corrections d’hallucinations pour construire des préférences d’entraînement et rapporte une baisse des hallucinations avec une meilleure qualité de réponse. [15]
« The fine-tuning through DPO leads to LLMs that exhibit a reduced rate of hallucinations and deliver improved answer quality. » [15]
- La prépublication non évaluée Finetune-RAG rapporte une amélioration de 21,2 % de l’exactitude factuelle par rapport au modèle de base, mais son entraînement simplifie les erreurs de recherche en associant exactement un fragment correct et un fragment incorrect. [18]
« Experimental results show that Finetune-RAG improves factual accuracy by 21.2% over the base model. » [18]
- RAGognizer rapporte qu’une supervision conjointe de génération et de détection au niveau des tokens réduit les hallucinations sans dégrader la qualité linguistique ou la pertinence, dans les conditions de cette seule étude. [26]
« Jointly optimizing language modeling and token-level hallucination detection consistently increases the separability of hallucinated and grounded tokens, while substantially reducing hallucination rates during generation without degrading language quality or relevance. » [26] · texte intégral
Évaluer la fidélité documentaire, la robustesse et les effets de présentation
Ces études distinguent l’exactitude de la réponse, son attribution aux documents et la capacité à résister à des preuves inadéquates. Elles montrent que disposer de documents pertinents ne garantit ni leur utilisation fidèle ni la stabilité de la réponse ; le papier 16 ne fournit toutefois, dans l’extrait disponible, aucun résultat exploitable.
- HaluEval montre que reconnaître les hallucinations reste difficile pour les modèles, mais que des connaissances externes ou des étapes de raisonnement peuvent améliorer cette reconnaissance, sans que cela équivaille à une mesure de génération sans hallucination. [4]
« However, our experiments also prove that providing external knowledge or adding reasoning steps can help LLMs recognize hallucinations. » [4]
- L’étude prépubliée sur les réponses longues constate une attribution seulement partielle aux documents, et le benchmark publié RGB révèle des difficultés à rejeter l’information insuffisante, intégrer les preuves et gérer de fausses informations. [7][8]
« 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. » [7]
1 autre phrase à l’appui
« Evaluation reveals that while LLMs exhibit a certain degree of noise robustness, they still struggle significantly in terms of negative rejection, information integration, and dealing with false information. » [8]
- Stable-RAG montre que permuter les mêmes documents modifie les réponses même lorsque le document de référence reste premier, tandis que l’étude Wikidata constate une dépendance de la fréquence mesurée des hallucinations à l’évaluateur. [30][27]
« We empirically show that under a Top-5 retrieval setting with the gold document included, LLM answers vary substantially across permutations of the retrieved set, even when the gold document is fixed in the first position. » [30]
1 autre phrase à l’appui
« This suggests that the observed frequency of hallucinations is influenced by the evaluation approach used. » [27]
Comprendre les hallucinations comme un problème de pipeline complet
Ces revues convergent vers une lecture systémique : les erreurs peuvent apparaître lors de la recherche, de l’intégration des sources ou de la génération. La revue générale de 2023 est une prépublication non évaluée ; les synthèses publiées ultérieures confortent cette lecture sans constituer autant de validations expérimentales indépendantes.
- Le bruit, les contradictions documentaires et la génération au-delà des preuves sont présentés comme des raisons pour lesquelles le RAG peut encore produire des hallucinations. [5][21]
« The presence of noise or contradictory information during retrieval can detrimentally affect RAG’s output quality. » [5] · texte intégral
1 autre phrase à l’appui
« The model may ignore retrieved evidence or generate beyond the evidence boundaries. » [21] · texte intégral
- Les revues organisent donc la mitigation autour de plusieurs étapes du pipeline, en incluant la détection et la correction plutôt qu’en assimilant l’ajout de recherche à une garantie de vérité. [19][23]
« We also investigate methods to reduce the impact of hallucination through detection and correction. » [19]
1 autre phrase à l’appui
« First, mitigation should be treated as a pipeline problem rather than a retrieval problem alone. » [23] · texte intégral
- Une revue systématique de 23 études sur les chatbots associe les gains des architectures avancées et modulaires au reclassement, au filtrage, à la vérification et à la génération attentive aux citations. [22]
« Findings show that advanced and modular RAG architectures generally improve factuality and faithfulness over purely parametric LLMs, especially when retrieval is paired with re-ranking, evidence filtering, verification, or citation-aware generation. » [22]
Où les articles s’accordent
- L’accès à des documents ne suffit pas : il faut à la fois retrouver des preuves pertinentes et en préserver fidèlement le contenu pendant la génération. [7][23]
2 phrases à l’appui
« LMs trained without retrieval and attribution in mind does not always generate sentences that can be attributed to in-context evidence documents, even when provided relevant documents only. » [7] · texte intégral
« Second, grounded generation requires both relevant evidence and faithful use of that evidence. » [23] · texte intégral
- La combinaison du RAG et d’un apprentissage adapté apparaît prometteuse, mais les résultats favorables ne constituent pas une garantie générale de sécurité ou d’absence d’hallucinations. [13][24]
Ce qu’ils laissent ouvert
- Il manque une comparaison harmonisée du RAG, du fine-tuning seul et de leur combinaison, à modèle, budget, corpus et mesure directe des hallucinations comparables ; les classements divergents disponibles ne constituent donc pas une contradiction expérimentale établie. [12][13][17][24]
Aucun article ne le dit en ces termes : déduit de ce que les articles n’étudient pas. Vérifiez-le.
- La transférabilité aux conditions réelles reste ouverte, notamment lorsque les tâches financières dépassent celles testées ou que de nombreux documents de qualité variable remplacent des paires synthétiques simples. [2][18]
2 phrases à l’appui
« Firstly, our tasks, although representative, cannot encompass the full spectrum of real-world tasks in the vast and varied domain of finance. » [2] · texte intégral
« This creates a simplified binary contrast that may not generalize to real-world scenarios where multiple retrieved documents vary in quality. » [18] · texte intégral
- L’évaluation devrait distinguer systématiquement exactitude, fidélité aux sources, attribution et pertinence de la recherche, car la standardisation actuelle reste limitée. [22][23]
2 phrases à l’appui
« However, mitigation remains constrained by retrieval noise, latency, limited benchmark standardization, and weak integration between hallucination detection and correction. » [22]
« Third, evaluation should move beyond answer accuracy toward factuality, faithfulness, attribution, and retrieval relevance. » [23] · texte intégral
- La généralisation des méthodes entraînées de détection aux interactions prolongées et à d’autres capacités du modèle reste insuffisamment établie. [26]
1 phrase à l’appui
« As a limitation of the current study, we focused on single-turn interactions and on one task (Q&A). » [26] · texte intégral
Références
-
[1] Retrieval Augmentation Reduces Hallucination in Conversation
Fondateur Récupération pour dialogue ancré · expérience
Pourquoi il est là : Évalue la récupération dans le dialogue et ses effets sur la connaissance et la qualité, sans comparaison directe au fine-tuning.
Résumé scientifique
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
-
[2] Deficiency of Large Language Models in Finance: An Empirical Examination of Hallucination
Préprint Évaluation de méthodes financières · expérience
Pourquoi il est là : Compare plusieurs méthodes de réduction des hallucinations en finance, dont le RAG, mais l’abstract ne précise pas une comparaison au fine-tuning.
Résumé scientifique · conclusions du texte intégral lues aussi
The hallucination issue is recognized as a fundamental deficiency of large language models (LLMs), especially when applied to fields such as finance, education, and law. Despite the growing concerns, there has been a lack of empirical investigation. In this paper, we provide an empirical examination of LLMs' hallucination behaviors in financial tasks. First, we empirically investigate LLM model's ability of explaining financial concepts and terminologies. Second, we assess LLM models' capacity of querying historical stock prices. Third, to alleviate the hallucination issue, we evaluate the efficacy of four practical methods, including few-shot learning, Decoding by Contrasting Layers (DoLa), the Retrieval Augmentation Generation (RAG) method and the prompt-based tool learning method for a function to generate a query command. Finally, our major finding is that off-the-shelf LLMs experience serious hallucination behaviors in financial tasks. Therefore, there is an urgent need to call for research efforts in mitigating LLMs' hallucination.
-
[3] Enhancing Retrieval-Augmented Large Language Models with Iterative Retrieval-Generation Synergy
récupération-génération itérative · expérience
Pourquoi il est là : Évalue une méthode de RAG itératif qui améliore la pertinence des preuves, mais ne la compare pas au fine-tuning seul.
Résumé scientifique
Retrieval-augmented generation has raise extensive attention as it is promising to address the limitations of large language models including outdated knowledge and hallucinations.However, retrievers struggle to capture relevance, especially for queries with complex information needs.Recent work has proposed to improve relevance modeling by having large language models actively involved in retrieval, i.e., to guide retrieval with generation.In this paper, we show that strong performance can be achieved by a method we call ITER-RETGEN, which synergizes retrieval and generation in an iterative manner: a model's response to a task input shows what might be needed to finish the task, and thus can serve as an informative context for retrieving more relevant knowledge which in turn helps generate a better response in another iteration.Compared with recent work which interleaves retrieval with generation when completing a single output, ITER-RETGEN processes all retrieved knowledge as a whole and largely preserves the flexibility in generation without structural constraints.We evaluate ITER-RETGEN on multi-hop question answering, fact verification, and commonsense reasoning, and show that it can flexibly leverage parametric knowledge and non-parametric knowledge, and is superior to or competitive with state-of-the-art retrieval-augmented baselines while causing fewer overheads of retrieval and generation.We can further improve performance via generation-augmented retrieval adaptation.
-
[4] HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Models
benchmark de détection d’hallucinations · méthode
Pourquoi il est là : Fournit un benchmark et des mesures de détection utiles à l’évaluation, mais ne compare pas RAG et fine-tuning.
Résumé scientifique
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.
-
[5] Retrieval-Augmented Generation for Large Language Models: A Survey
Revue Préprint revue des architectures RAG · revue
Pourquoi il est là : Présente les variantes et composants du RAG pertinents pour expliquer les différences d’effet, mais pas une comparaison avec le fine-tuning.
Résumé scientifique · conclusions du texte intégral lues aussi
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.
-
[6] Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection
Préprint Récupération adaptative et auto-critique · méthode
Pourquoi il est là : Propose un RAG avec récupération à la demande et auto-évaluation, éclairant les mécanismes de grounding, sans comparaison au fine-tuning seul.
Résumé scientifique
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.
-
[7] Understanding Retrieval Augmentation for Long-Form Question Answering
Préprint Études contrôlées de réponses longues · expérience
Pourquoi il est là : Analyse l’attribution des réponses longues aux documents récupérés, une mesure importante du grounding, sans comparer au fine-tuning seul.
Résumé scientifique · conclusions du texte intégral lues aussi
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.
-
[8] Benchmarking Large Language Models in Retrieval-Augmented Generation
Récent benchmark des capacités RAG · méthode
Pourquoi il est là : Le benchmark mesure la robustesse, le rejet et l’intégration d’information en RAG, éclairant la dépendance des résultats aux capacités évaluées.
Résumé scientifique
Retrieval-Augmented Generation (RAG) is a promising approach for mitigating the hallucination of large language models (LLMs). However, existing research lacks rigorous evaluation of the impact of retrieval-augmented generation on different large language models, which make it challenging to identify the potential bottlenecks in the capabilities of RAG for different LLMs. In this paper, we systematically investigate the impact of Retrieval-Augmented Generation on large language models. We analyze the performance of different large language models in 4 fundamental abilities required for RAG, including noise robustness, negative rejection, information integration, and counterfactual robustness. To this end, we establish Retrieval-Augmented Generation Benchmark (RGB), a new corpus for RAG evaluation in both English and Chinese. RGB divides the instances within the benchmark into 4 separate testbeds based on the aforementioned fundamental abilities required to resolve the case. Then we evaluate 6 representative LLMs on RGB to diagnose the challenges of current LLMs when applying RAG. Evaluation reveals that while LLMs exhibit a certain degree of noise robustness, they still struggle significantly in terms of negative rejection, information integration, and dealing with false information. The aforementioned assessment outcomes indicate that there is still a considerable journey ahead to effectively apply RAG to LLMs.
-
[9] Benchmarking Retrieval-Augmented Generation for Medicine
Récent benchmark médical du RAG · méthode
Pourquoi il est là : Le benchmark compare de nombreuses combinaisons de corpus, récupérateurs et modèles en médecine, montrant l’importance des choix d’évaluation.
Résumé scientifique
While large language models (LLMs) have achieved state-of-the-art performance on a wide range of medical question answering (QA) tasks, they still face challenges with hallucinations and outdated knowledge.Retrievalaugmented generation (RAG) is a promising solution and has been widely adopted.However, a RAG system can involve multiple flexible components, and there is a lack of best practices regarding the optimal RAG setting for various medical purposes.To systematically evaluate such systems, we propose the Medical Information Retrieval-Augmented Generation Evaluation (MIRAGE), a first-of-its-kind benchmark including 7,663 questions from five medical QA datasets.Using MIRAGE, we conducted large-scale experiments with over 1.8 trillion prompt tokens on 41 combinations of different corpora, retrievers, and backbone LLMs through the MEDRAG toolkit introduced in this work.Overall, MEDRAG improves the accuracy of six different LLMs by up to 18% over chain-of-thought prompting, elevating the performance of GPT-3.5 and Mixtral to GPT-4level.Our results show that the combination of various medical corpora and retrievers achieves the best performance.In addition, we discovered a log-linear scaling property and the "lostin-the-middle" effects in medical RAG.We believe our comprehensive evaluations can serve as practical guidelines for implementing RAG systems for medicine 1 .
-
[10] Chain-of-Note: Enhancing Robustness in Retrieval-Augmented Language Models
Récent Notes de lecture séquentielles · méthode
Pourquoi il est là : Propose une méthode pour limiter les erreurs dues aux documents non pertinents et gérer l’incertitude en RAG.
Résumé scientifique
Retrieval-augmented language model (RALM) represents a significant advancement in mitigating factual hallucination by leveraging external knowledge sources.However, the reliability of the retrieved information is not always guaranteed, and the retrieval of irrelevant data can mislead the response generation.Moreover, standard RALMs frequently neglect their intrinsic knowledge due to the interference from retrieved information.In instances where the retrieved information is irrelevant, RALMs should ideally utilize their intrinsic knowledge or, in the absence of both intrinsic and retrieved knowledge, opt to respond with "unknown" to avoid hallucination.In this paper, we introduces CHAIN-OF-NOTE (CON), a novel approach to improve robustness of RALMs in facing noisy, irrelevant documents and in handling unknown scenarios.The core idea of CON is to generate sequential reading notes for each retrieved document, enabling a thorough evaluation of their relevance to the given question and integrating this information to formulate the final answer.Our experimental results show that GPT-4, when equipped with CON, outperforms the CHAIN-OF-THOUGHT approach.Besides, we utilized GPT-4 to create 10K CON data, subsequently trained on LLaMa-2 7B model.Our experiments across four open-domain QA benchmarks show that fine-tuned RALMs equipped with CON significantly outperform standard fine-tuned RALMs.
-
Récent Préprint RAG et fine-tuning spécialisé · expérience
Pourquoi il est là : Évalue un système combinant RAG et fine-tuning pour des requêtes privées et temporelles, et discute les limites du fine-tuning seul.
Résumé scientifique · conclusions du texte intégral lues aussi
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.
-
[12] Fine-Tuning or Retrieval? Comparing Knowledge Injection in LLMs
Récent Comparaison RAG et fine-tuning · expérience
Pourquoi il est là : Compare directement le RAG et le fine-tuning non supervisé sur plusieurs tâches intensives en connaissances.
Résumé scientifique
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.
-
Récent Préprint fine-tuning et variantes RAG · expérience
Pourquoi il est là : Évalue conjointement le fine-tuning de petits modèles et plusieurs stratégies RAG, et indique que le RAG seul ne suffit pas toujours.
Résumé scientifique · conclusions du texte intégral lues aussi
Hallucination is a key roadblock for applications of Large Language Models (LLMs), particularly for enterprise applications that are sensitive to information accuracy. To address this issue, two general approaches have been explored: Retrieval-Augmented Generation (RAG) to supply LLMs with updated information as context, and fine-tuning the LLMs with new information and desired output styles. In this paper, we propose Honest AI: a novel strategy to fine-tune "small" language models to say "I don't know" to reduce hallucination, along with several alternative RAG approaches. The solution ranked 1st in Task 2 for the false premise question. The alternative approaches include using RAG with search engine and knowledge graph results, fine-tuning base LLMs with new information and combinations of both approaches. Although all approaches improve the performance of the LLMs, RAG alone does not significantly improve the performance and fine-tuning is needed for better results. Finally, the hybrid approach achieved the highest score in the CRAG benchmark. In addition, our approach emphasizes the use of relatively small models with fewer than 10 billion parameters, promoting resource efficiency.
-
Récent RAG avec correcteur adaptatif · expérience
Pourquoi il est là : Évalue un RAG corrigé pour la description spatiale visuelle et rapporte moins d’hallucinations, sans comparaison au fine-tuning seul.
Résumé scientifique
Visual Spatial Description (VSD) is an emerging image-to-text task which aims at generating descriptions of the spatial relationships between given objects in an image. In this paper, we apply Retrieval-Augmented Generation (RAG) technology in guiding Multimodal Large Language Models (MLLMs) for the task of VSD, complemented by an Adaptive Hallucination Corrector, and further fine-tuning them to bolster semantic understanding and overall model efficacy. We found that our approach demonstrated higher accuracy and fewer hallucination errors in both spatial relationship classification and visual language description tasks within the VSD task, achieving state-of-the-art results.
-
[15] RAG-HAT: A Hallucination-Aware Tuning Pipeline for LLM in Retrieval-Augmented Generation
Récent ajustement sensible aux hallucinations · méthode
Pourquoi il est là : Décrit une chaîne de détection et de correction des hallucinations en RAG, sans comparaison explicite au fine-tuning seul.
Résumé scientifique
Retrieval-augmented generation (RAG) has emerged as a significant advancement in the field of large language models (LLMs).By integrating up-to-date information not available during their initial training, RAG greatly enhances the practical utility of LLMs in realworld applications.However, even with RAG, LLMs can still produce inaccurate outputs, such as distorting or misinterpreting source content, posing risks in high-trust scenarios.To address these issues, we introduce a novel approach called Hallucination Aware Tuning (HAT).This method involves training hallucination detection models that generate detection labels and provide detailed descriptions of the detected hallucinations.Utilizing these detection results-particularly the hallucination descriptions-GPT-4 Turbo is employed to correct any detected hallucinations.The corrected outputs, free of hallucinations, along with the original versions, are used to create a preference dataset for Direct Preference Optimization (DPO) training.The fine-tuning through DPO leads to LLMs that exhibit a reduced rate of hallucinations and deliver improved answer quality.
-
Récent Évaluation des effets de récupération · expérience
Pourquoi il est là : Étudie quand la récupération aide ou nuit aux modèles de langage, éclairant les conditions d’évaluation, sans comparer au fine-tuning seul.
Résumé scientifique
Seiji Maekawa, Hayate Iso, Sairam Gurajada, Nikita Bhutani. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.
-
[17] Exploring RAG Solutions to Reduce Hallucinations in LLMs
Récent comparaison de systèmes RAG · expérience
Pourquoi il est là : Compare RAG naïf et Graph RAG à des modèles de base et un modèle fine-tuné sur plusieurs jeux de données.
Résumé scientifique
Large Language Models (LLMs) often face challenges in generating accurate and reliable information, particularly in knowledge-intensive tasks. This limitation, referred to as hallucination, occurs when models produce content that is incorrect, irrelevant, or unsupported by evidence. Retrieval Augmented Generation (RAG) solutions provide a promising approach by integrating relevant external knowledge, enabling models to generate factually grounded responses. This study evaluates the performance of a base LLM model, a fine-tuned DistilBERT model, and two RAG architectures, Naïve RAG and Graph RAG, to study their impact on reducing hallucinations and enhancing contextual understanding. Using subsets of HaluEval, Squad-V2, and TriviaQA benchmark datasets, the base model achieved accuracies of 10.18%, 12.67%, and 5.46% respectively; Naive RAG resulted in 44.56%, 19.04%, and 35.32% accuracies; while the fine-tuned LLM model's accuracies were 72.5%, 72.31%, and 88.7% respectively. Graph RAG resulted in 8.85% and 15.12% accuracies using Squad-V2 and TriviaQA, respectively. Our findings show that while fine-tuned LLMs outperform baseline models, incorporating RAG solutions did not result in significant performance improvements, suggesting that the incorporation of external knowledge may not always align with the needs of the task. Experiments demonstrate that Graph RAG handles complex queries by leveraging relationships within structured knowledge graphs. Data organized as a knowledge graph may enable Graph RAG solutions reach their full potential by utilizing their capacity to efficiently retrieve contextually relevant information. Although computing complexity remains a restriction, this study shows that RAG topologies might not consistently enhance LLM reliability in practical situations.
-
[18] Finetune-RAG: Fine-Tuning Language Models to Resist Hallucination in Retrieval-Augmented Generation
Récent Préprint fine-tuning pour RAG imparfait · expérience
Pourquoi il est là : Compare une méthode de fine-tuning conçue pour les erreurs de récupération à un modèle de base et rapporte un gain de précision factuelle.
Résumé scientifique · conclusions du texte intégral lues aussi
Retrieval-Augmented Generation (RAG) has emerged as a powerful framework to improve factuality in large language models (LLMs) by grounding their outputs in retrieved documents. However, ensuring perfect retrieval of relevant information remains challenging, and when irrelevant content is passed downstream to an LLM, it can lead to hallucinations. In this work, we propose Finetune-RAG, a simple and effective fine-tuning approach that features the first-of-its-kind RAG training dataset constructed to mimic real-world imperfections. Experimental results show that Finetune-RAG improves factual accuracy by 21.2% over the base model. We also propose Bench-RAG, an LLM-as-a-judge evaluation pipeline that stress tests models under realistic imperfect retrieval scenarios. Our codebase and dataset are fully open sourced for community use.
-
[19] Hallucination Mitigation for Retrieval-Augmented Large Language Models: A Review
Récent Revue revue des hallucinations en RAG · revue
Pourquoi il est là : Passe en revue les causes et mesures d’atténuation propres au RAG, sans comparaison directe avec le fine-tuning seul.
Résumé scientifique
Retrieval-augmented generation (RAG) leverages the strengths of information retrieval and generative models to enhance the handling of real-time and domain-specific knowledge. Despite its advantages, limitations within RAG components may cause hallucinations, or more precisely termed confabulations in generated outputs, driving extensive research to address these limitations and mitigate hallucinations. This review focuses on hallucination in retrieval-augmented large language models (LLMs). We first examine the causes of hallucinations from different sub-tasks in the retrieval and generation phases. Then, we provide a comprehensive overview of corresponding hallucination mitigation techniques, offering a targeted and complete framework for addressing hallucinations in retrieval-augmented LLMs. We also investigate methods to reduce the impact of hallucination through detection and correction. Finally, we discuss promising future research directions for mitigating hallucinations in retrieval-augmented LLMs.
-
Récent Revue Méta-analyse du RAG biomédical · revue
Pourquoi il est là : Synthétise et quantifie les comparaisons empiriques entre performances LLM de base et RAG en biomédecine.
Résumé scientifique · conclusions du texte intégral lues aussi
OBJECTIVE: The objectives of this study are to synthesize findings from recent research of retrieval-augmented generation (RAG) and large language models (LLMs) in biomedicine and provide clinical development guidelines to improve effectiveness. MATERIALS AND METHODS: We conducted a systematic literature review and a meta-analysis. The report was created in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 analysis. Searches were performed in 3 databases (PubMed, Embase, PsycINFO) using terms related to "retrieval augmented generation" and "large language model," for articles published in 2023 and 2024. We selected studies that compared baseline LLM performance with RAG performance. We developed a random-effect meta-analysis model, using odds ratio as the effect size. RESULTS: Among 335 studies, 20 were included in this literature review. The pooled effect size was 1.35, with a 95% confidence interval of 1.19-1.53, indicating a statistically significant effect (P = .001). We reported clinical tasks, baseline LLMs, retrieval sources and strategies, as well as evaluation methods. DISCUSSION: Building on our literature review, we developed Guidelines for Unified Implementation and Development of Enhanced LLM Applications with RAG in Clinical Settings to inform clinical applications using RAG. CONCLUSION: Overall, RAG implementation showed a 1.35 odds ratio increase in performance compared to baseline LLMs. Future research should focus on (1) system-level enhancement: the combination of RAG and agent, (2) knowledge-level enhancement: deep integration of knowledge into LLM, and (3) integration-level enhancement: integrating RAG systems within electronic health records.
-
[21] Retrieval-Augmented Generation and Hallucination in Large Language Models: A Scholarly Overview
Récent Revue revue de RAG anti-hallucination · revue
Pourquoi il est là : Présente les mécanismes et l’efficacité du RAG, mais l’abstract ne compare pas directement avec le fine-tuning.
Résumé scientifique · conclusions du texte intégral lues aussi
Large Language Models (LLMs) have revolutionized natural language processing tasks, yet they often suffer from "hallucination” the confident generation of factually incorrect information. Retrieval-Augmented Generation (RAG) has emerged as a promising technique to mitigate hallucinations by grounding model responses in external documents. This article explores the underlying causes of hallucinations in LLMs, the mechanisms and architectures of RAG systems, their effectiveness in reducing hallucinations, and ongoing challenges. We conclude with a discussion of future directions for integrating retrieval mechanisms more seamlessly into generative architecture.
-
Récent Revue revue systématique du RAG · revue
Pourquoi il est là : Synthétise architectures, récupération et évaluations du RAG en chatbots, mais l’abstract ne rapporte pas de comparaison au fine-tuning seul.
Résumé scientifique
Large language model (LLM) chatbots are increasingly deployed in education, health care, and decision-support settings, yet their reliability is limited by hallucination: fluent responses that are factually incorrect, weakly grounded, or inconsistent with supplied evidence. Retrieval-Augmented Generation (RAG) addresses this limitation by conditioning generation on external knowledge retrieved at inference time. This paper presents a systematic review of RAG techniques for hallucination mitigation in chatbot systems. Following the PRISMA 2020 process, studies published from 2020 to 2026 were searched in Scopus, IEEE Xplore, and Google Scholar. After duplicate removal, screening, and full-text eligibility assessment, 23 studies were synthesized qualitatively. The review analyzes RAG architectures, retrieval strategies, evidence-integration mechanisms, and evaluation practices. Findings show that advanced and modular RAG architectures generally improve factuality and faithfulness over purely parametric LLMs, especially when retrieval is paired with re-ranking, evidence filtering, verification, or citation-aware generation. However, mitigation remains constrained by retrieval noise, latency, limited benchmark standardization, and weak integration between hallucination detection and correction. The paper concludes with research directions for trustworthy, efficient, and explainable RAG-based chatbot systems.
-
Récent Revue revue des mécanismes et évaluations · revue
Pourquoi il est là : Analyse les causes, atténuations et évaluations des hallucinations en RAG, utiles pour interpréter les différences selon les réglages.
Résumé scientifique · conclusions du texte intégral lues aussi
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.
-
Récent Revue Revue des hybrides santé · revue
Pourquoi il est là : Synthétise les études empiriques combinant fine-tuning et RAG en santé, directement pertinentes pour leurs effets complémentaires.
Résumé scientifique · conclusions du texte intégral lues aussi
(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.
-
[25] Mitigating Hallucination on Hallucination in RAG via Ensemble Voting
Récent vote multi-agents pour RAG · expérience
Pourquoi il est là : Compare empiriquement une méthode de RAG aux hallucinations induites par de mauvaises récupérations, sans référence au fine-tuning seul.
Résumé scientifique · conclusions du texte intégral lues aussi
Retrieval-Augmented Generation (RAG) aims to reduce hallucinations in Large Language Models (LLMs) by integrating external knowledge. However, RAG introduces a critical challenge: hallucination on hallucination," where flawed retrieval results mislead the generation model, leading to compounded hallucinations. To address this issue, we propose VOTE-RAG, a novel, training-free framework with a two-stage structure and efficient, parallelizable voting mechanisms. VOTE-RAG includes: (1) Retrieval Voting, where multiple agents generate diverse queries in parallel and aggregate all retrieved documents; (2) Response Voting, where multiple agents independently generate answers based on the aggregated documents, with the final output determined by majority vote. We conduct comparative experiments on six benchmark datasets. Our results show that VOTE-RAG achieves performance comparable to or surpassing more complex frameworks. Additionally, VOTE-RAG features a simpler architecture, is fully parallelizable, and avoids the problem drift" risk. Our work demonstrates that simple, reliable ensemble voting is a superior and more efficient method for mitigating RAG hallucinations.
-
[26] RAGognizer: Hallucination-Aware Fine-Tuning via Detection Head Integration
Récent fine-tuning avec détection intégrée · méthode
Pourquoi il est là : Propose un fine-tuning sensible aux hallucinations en contexte RAG, utile pour comprendre l’interaction des approches, mais sans comparaison directe annoncée.
Résumé scientifique · conclusions du texte intégral lues aussi
Retrieval-Augmented Generation (RAG) is widely used to augment the input to Large Language Models (LLMs) with external information, such as recent or domain-specific knowledge. Nonetheless, current models still produce closed-domain hallucinations and generate content that is unsupported by the retrieved context. Current detection approaches typically treat hallucination as a post-hoc problem, relying on black-box consistency checks or probes over frozen internal representations. In this work, we demonstrate that hallucination detection based on internal state representation can also serve as a direct training signal. We introduce RAGognize, a dataset of naturally occurring closed-domain hallucinations with token-level annotations, and RAGognizer, a hallucination-aware fine-tuning approach that integrates a lightweight detection head into an LLM, allowing for the joint optimization of language modeling and hallucination detection. This joint objective forces the model to improve the separability of its internal states regarding hallucinations while simultaneously learning to generate well-formed and meaningful responses. Across multiple benchmarks, RAGognizer achieves state-of-the-art token-level hallucination detection while substantially reducing hallucination rates during generation, without degrading language quality or relevance.
-
Récent RAG structuré avec Wikidata · expérience
Pourquoi il est là : Évalue empiriquement le RAG sur des questions multi-sauts avec plusieurs mesures d’hallucination, mais sans comparaison au fine-tuning.
Résumé scientifique
Hallucinations continue to significantly restrict large language models (LLMs), particularly in fact-finding and knowledge-intensive tasks. Retrieval-Augmented Generation (RAG), which bases model outputs on external knowledge sources, addresses this problem. This study examines the impact of integrating structured knowledge from Wikidata into an LLM-based question-answering pipeline to reduce hallucinated material. The proposed Wikidata-based RAG technique injects this structured context into the LLM prompt before response synthesis by first extracting entities from the input question and then gathering candidate entities and verified attributes from Wikidata. Both the prompt-only baseline models and the Wikidata-grounded RAG system were used to evaluate 80 multi-hop compositional problems. The proposed RAG model is evaluated using four hallucination evaluation methods. The model is evaluated against three prompt-only baselines: the AimonLabs hallucination-detection model, the Vectara HHEM, and an LLM-as-a-Judge rubric. Retrieved-Context Faithfulness (RAGTruth-style evaluation) is assessed using the last evaluation method. (Vectara HHEM: 77.5% vs 38.75% factual replies; AIMON: 60% vs 32.5% factual responses; LLM-as-a-Judge: 75% vs 38.75% factual responses) RAG consistently surpasses the baseline across all four evaluators. 54 of 79 assessed examples had a retrieved-context fidelity of 68.35%. This suggests that the observed frequency of hallucinations is influenced by the evaluation approach used. These results show how fact accuracy may be greatly increased by basing LLMs on structured knowledge from Wikidata. They also show how important it is to have several assessors when studying hallucinations.
-
Récent Revue Préprint revue comparative en médecine · revue
Pourquoi il est là : Compare explicitement RAG et fine-tuning dans des études médicales et rapporte leur efficacité relative contre les hallucinations.
Résumé scientifique
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.
-
Récent Revue comparaison de méthodes RAG · revue
Pourquoi il est là : Examine plusieurs variantes de RAG anti-hallucination, mais pas leur comparaison avec le fine-tuning seul.
Résumé scientifique
Although large language models have shown strong ability in generating text, they can still give answers that look fluent but are factually incorrect or lack clear support, especially when the questions are related to factual knowledge, specific fields, or information that changes quickly. This issue is commonly called hallucination. Retrieval-Augmented Generation (RAG), provides a useful way to reduce this problem, because it lets the model use external documents before producing the final answer. This paper mainly discusses three representative RAG-based methods for reducing hallucinations, including Basic RAG, Self-RAG, and Corrective RAG. Basic RAG uses a simple process of retrieving first and then generating, so the generated content can be more closely connected with factual information. Self-RAG adds a self-reflection process to this basic structure, and the model can decide whether retrieval is necessary, as well as whether the answer has enough evidence support. Corrective RAG pays more attention to the quality of retrieved information; it checks and corrects poor retrieved content before the generation stage. According to their working processes, limitations, and related experimental results, this paper also compares these three methods from several aspects, such as factual reliability, retrieval control, use of evidence, system complexity, and suitable application scenarios.
-
Récent robustesse aux permutations documentaires · expérience
Pourquoi il est là : Montre que l’ordre des documents récupérés influe sur les hallucinations, un facteur important des résultats RAG selon les réglages.
Résumé scientifique
Retrieval-Augmented Generation (RAG) has become a key paradigm for reducing factual hallucinations in Large Language Models (LLMs), yet little is known about how the order of retrieved documents affects model behavior.We empirically show that under a Top-5 retrieval setting with the gold document included, LLM answers vary substantially across permutations of the retrieved set, even when the gold document is fixed in the first position.This reveals a previously underexplored sensitivity to retrieval permutations.Although existing robust RAG methods focus primarily on enhancing LLM robustness to low-quality retrieval and mitigating positional bias to distribute attention fairly over long contexts, neither approach directly addresses permutation sensitivity.In this paper, we propose Stable-RAG, which exploits permutation sensitivity estimation to mitigate permutation-induced hallucinations.Stable-RAG runs the generator under multiple retrieval orders, clusters hidden states, and decodes from a cluster-center representation that captures the dominant reasoning pattern.It then uses these reasoning results to align hallucinated outputs toward the correct answer, encouraging the model to produce consistent and accurate predictions across document permutations.Experiments on three QA datasets show that Stable-RAG improves answer accuracy, reasoning consistency, and generalization across datasets, retrievers, and input lengths compared with strong baselines 1 .