Academic literature on the topic 'Large Language Models (LLM)'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the lists of relevant articles, books, theses, conference reports, and other scholarly sources on the topic 'Large Language Models (LLM).'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Journal articles on the topic "Large Language Models (LLM)"
1
Fang, Meng, Shilong Deng, Yudi Zhang, Zijing Shi, Ling Chen, Mykola Pechenizkiy, and Jun Wang. "Large Language Models Are Neurosymbolic Reasoners." Proceedings of the AAAI Conference on Artificial Intelligence 38, no. 16 (March 24, 2024): 17985–93. http://dx.doi.org/10.1609/aaai.v38i16.29754.
Full textAbstract:
A wide range of real-world applications is characterized by their symbolic nature, necessitating a strong capability for symbolic reasoning. This paper investigates the potential application of Large Language Models (LLMs) as symbolic reasoners. We focus on text-based games, significant benchmarks for agents with natural language capabilities, particularly in symbolic tasks like math, map reading, sorting, and applying common sense in text-based worlds. To facilitate these agents, we propose an LLM agent designed to tackle symbolic challenges and achieve in-game objectives. We begin by initializing the LLM agent and informing it of its role. The agent then receives observations and a set of valid actions from the text-based games, along with a specific symbolic module. With these inputs, the LLM agent chooses an action and interacts with the game environments. Our experimental results demonstrate that our method significantly enhances the capability of LLMs as automated agents for symbolic reasoning, and our LLM agent is effective in text-based games involving symbolic tasks, achieving an average performance of 88% across all tasks.
2
Wang, Runze, Mingqi Yang, and Yanming Shen. "Bridging Molecular Graphs and Large Language Models." Proceedings of the AAAI Conference on Artificial Intelligence 39, no. 20 (April 11, 2025): 21234–42. https://doi.org/10.1609/aaai.v39i20.35422.
Full textAbstract:
While Large Language Models (LLMs) have shown exceptional generalization capabilities, their ability to process graph data, such as molecular structures, remains limited. To bridge this gap, this paper proposes Graph2Token, an efficient solution that aligns graph tokens to LLM tokens. The key idea is to represent a graph token with the LLM token vocabulary, without fine-tuning the LLM backbone. To achieve this goal, we first construct a molecule-text paired dataset from multi-sources, including CHEBI and HMDB, to train a graph structure encoder, which reduces the distance between graphs and texts representations in the feature space. Then, we propose a novel alignment strategy that associates a graph token with LLM tokens. To further unleash the potential of LLMs, we collect molecular IUPAC name identifiers, which are incorporated into the LLM prompts. By aligning molecular graphs as special tokens, we can activate LLMs' generalization ability to molecular few-shot learning. Extensive experiments on molecular classification and regression tasks demonstrate the effectiveness of our proposed Graph2Token.
3
Mochihashi, Daichi. "Large Language Models（LLM）and Robotics." Journal of the Robotics Society of Japan 40, no. 10 (2022): 863–66. http://dx.doi.org/10.7210/jrsj.40.863.
Full text
4
Devyatkin, Dmitry A., Vladimir A. Salimovsky, Natalia V. Chudova, Anastasia A. Ryzhova, and Oleg G. Grigoriev. "Large language models and speech genre systematicity." International Journal “Speech Genres” 20, no. 1 (45) (February 21, 2025): 6–23. https://doi.org/10.18500/2311-0740-2025-20-1-45-6-23.
Full textAbstract:
The paper examines a large language model (LLM) to recognize speech genres. Although artificial neural networks are effectively utilized in many important fields, they, however, have a serious drawback. The mechanism of their functioning is hidden from researchers; therefore, the results of their use do not get explanation. The purpose of the study is to reveal the basic mechanisms of functioning of the linguistic model LLM (Transformer) and thereby ensure the interpretability of the data it provides. The research is based on two genres of academic text: “Description of a new scientific phenomenon” and “Explication of a scientific concept.” We verified a hypothesis according to which the LLM feature set is based on the speech systematicity of the recognized genres. It is also shown that since genre-speech systematicity is determined by extralinguistic factors, primarily the characteristics of human consciousness, its manifestations, reflected in the hidden state of the LLM, can be used to model cognitive processes embodied in speech. We also analyze existing approaches to the interpretation of LLMs and describe the applied method to do it. The paper provides the following linguistic interpretation of LLM training and fine-tuning: preliminary training on large text corpora allows a model to display language resources (a system of linguistic units and general principles of their use) relatively completely, while fine-tuning on samples of a certain genre-speech organization restructures the linguistic systematicity into speech systematicity. During the experiments we decoded the hidden state of the LLM and accurately reproduced the composition and frequency of lexis from the training dataset. The classification score for each of the considered genres by the LLM is F1 0.99, we believe this is because of their speech consistency.
5
Yang, Jidong. "Large language models privacy and security." Applied and Computational Engineering 76, no. 1 (July 16, 2024): 177–88. http://dx.doi.org/10.54254/2755-2721/76/20240584.
Full textAbstract:
The advancement of large language models (LLMs) has yielded significant advancements across various domains. Nevertheless, this progress has also raised crucial concerns regarding privacy and security. The paper does a comprehensive literature study to thoroughly examine the fundamental principles of LLM. It also provides a detailed examination of the characteristics and application fields of various LLMs, with a particular focus on Transformer. Furthermore, this study places emphasis on the examination of privacy concerns that may emerge in the context of LLM's handling of personal and sensitive data. It also explores the potential hazards associated with information leakage and misuse, as well as the existing privacy safeguards and the obstacles encountered in their implementation. Overall, LLM has made significant advancements in technology. However, it is imperative to acknowledge the importance of doing research on safeguarding privacy and enhancing security. These aspects are vital for guaranteeing the sustained development and public confidence in LLM technology.
6
Shanahan, Murray. "Talking about Large Language Models." Communications of the ACM 67, no. 2 (January 25, 2024): 68–79. http://dx.doi.org/10.1145/3624724.
Full textAbstract:
Interacting with a contemporary LLM-based conversational agent can create an illusion of being in the presence of a thinking creature. Yet, in their very nature, such systems are fundamentally not like us.
7
Liu, Yuxin. "Attention is All Large Language Model Need." ITM Web of Conferences 73 (2025): 02025. https://doi.org/10.1051/itmconf/20257302025.
Full textAbstract:
With the advent of the Transformer, the attention mechanism has been applied to Large Language Model (LLM), evolving from initial single- modal large models to today's multi-modal large models. This has greatly propelled the development of Artificial Intelligence (AI) and ushered humans into the era of large models. Single-modal large models can be broadly categorized into three types based on their application domains: Text LLM for Natural Language Processing (NLP), Image LLM for Computer Vision (CV), and Audio LLM for speech interaction. Multi-modal large models, on the other hand, can leverage multiple data sources simultaneously to optimize the model. This article also introduces the training process of the GPT series. Large models have also had a significant impact on industry and society, bringing with them a number of unresolved problems. The purpose of this article is to assist researchers in comprehending the various forms of LLM, as well as its development, pre- training architecture, difficulties, and future objectives.
8
Ma, Ziyang, Guanrou Yang, Yifan Yang, Zhifu Gao, Jiaming Wang, Zhihao Du, Fan Yu, et al. "Speech Recognition Meets Large Language Model: Benchmarking, Models, and Exploration." Proceedings of the AAAI Conference on Artificial Intelligence 39, no. 23 (April 11, 2025): 24840–48. https://doi.org/10.1609/aaai.v39i23.34666.
Full textAbstract:
In this paper, we focus on prompting one of the most important tasks in the field of speech processing, i.e., automatic speech recognition (ASR), with speech foundation encoders and large language models (LLM). Despite the growing body of research in this area, we find that many crucial design decisions in LLM-based ASR systems are often inadequately justified. This lack of clarity impedes the field's progress, making it challenging to pinpoint which design choices truly improve model performance. To address these challenges, we conduct a comprehensive series of experiments that explore various aspects, leading to the optimal LLM-based ASR system. We found that delicate designs are not necessary, while a clean setup with little task-specific design is competent. The models achieve strong performance on the Librispeech and Gigaspeech datasets, compared to both LLM-based models and non-LLM-based models. Finally, we explore the capability emergence of LLM-based ASR in the process of modal alignment. We hope that our study can facilitate the research on extending LLM with cross-modality capacity and shed light on the LLM-based ASR community.
9
Zelenkov, Yuri A. "Knowledge management in organization and the large language models." Russian Management Journal 22, no. 3 (2024): 573–601. https://doi.org/10.21638/spbu18.2024.309.
Full textAbstract:
Purpose: to summarize, classify and analyze current academic papers on the use of large language models (LLM) in knowledge management in organization. Methodology: systematic literature review was conducted. It was based on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) framework. 75 papers were selected for the analysis, including academic papers and reports of consulting companies published since 2020. Findings: four main research areas have been identified: (1) LLM implementation issues; (2) the impact of LLM on knowledge management efficiency; the application of LLM in the processes of (3) knowledge usage and (4) knowledge creation. Within each area, the key papers and open questions have been reviewed. Originality and contribution: the paper presents a systematic review of current publications, proposes a classification of research topics, and identifies potential directions for new research. The study also considers limitations hindering the implementation of LLM in the organization's knowledge management practice.
10
Martínez, Gonzalo, Javier Conde, Elena Merino-Gómez, Beatriz Bermúdez-Margaretto, José Alberto Hernández, Pedro Reviriego, and Marc Brysbaert. "Establishing vocabulary tests as a benchmark for evaluating large language models." PLOS ONE 19, no. 12 (December 12, 2024): e0308259. https://doi.org/10.1371/journal.pone.0308259.
Full textAbstract:
Vocabulary tests, once a cornerstone of language modeling evaluation, have been largely overlooked in the current landscape of Large Language Models (LLMs) like Llama 2, Mistral, and GPT. While most LLM evaluation benchmarks focus on specific tasks or domain-specific knowledge, they often neglect the fundamental linguistic aspects of language understanding. In this paper, we advocate for the revival of vocabulary tests as a valuable tool for assessing LLM performance. We evaluate seven LLMs using two vocabulary test formats across two languages and uncover surprising gaps in their lexical knowledge. These findings shed light on the intricacies of LLM word representations, their learning mechanisms, and performance variations across models and languages. Moreover, the ability to automatically generate and perform vocabulary tests offers new opportunities to expand the approach and provide a more complete picture of LLMs’ language skills.
Dissertations / Theses on the topic "Large Language Models (LLM)"
1
Naqvi, Syed Muhammad Raza. "Exploration des LLM et de l'XAI sémantique pour les capacités des robots industriels et les connaissances communes en matière de fabrication." Electronic Thesis or Diss., Université de Toulouse (2023-....), 2025. http://www.theses.fr/2025TLSEP014.
Full textAbstract:
Dans l'industrie 4.0, la fabrication avancée est essentielle pour façonner les usines du futur, en permettant d'améliorer la planification, l'ordonnancement et le contrôle. La capacité d'adapter rapidement les lignes de production en réponse aux demandes des clients ou à des situations inattendues est essentielle pour améliorer l'avenir de la fabrication. Bien que l'IA apparaisse comme une solution, les industries s'appuient toujours sur l'expertise humaine en raison des problèmes de confiance et du manque de transparence des décisions de l'IA. L'IA explicable intégrant des connaissances de base liées à la fabrication est cruciale pour rendre les décisions de l'IA compréhensibles et dignes de confiance. Dans ce contexte, nous proposons le cadre S-XAI, une solution intégrée combinant les spécifications de la machine et le MCSK pour fournir une prise de décision explicable et transparente. L'accent est mis sur la fourniture de capacités machine en temps réel afin de garantir une prise de décision précise tout en expliquant simultanément le processus de prise de décision à toutes les parties prenantes concernées. En conséquence, le premier objectif était de formaliser les spécifications des machines, y compris les capacités, les fonctions, la qualité et les caractéristiques des processus, en se concentrant sur la robotique. Pour ce faire, nous avons créé une ontologie des capacités des robots qui formalise tous les aspects pertinents des spécifications des machines, tels que la capacité, l'aptitude, la fonction, la qualité et les caractéristiques du processus. En plus de cette formalisation, le RCO permet aux acteurs de la fabrication de capturer les capacités robotiques décrites dans les manuels de spécification (capacités annoncées) et de les comparer avec les performances réelles (capacités opérationnelles). Le RCO est basé sur le langage de description des services de machines, une ontologie de référence créée pour les services de fabrication et alignée sur l'ontologie formelle de base, l'ontologie de la fonderie industrielle, l'ontologie des artefacts d'information et l'ontologie des relations. Le deuxième objectif était la formalisation du MCSK. Nous introduisons le MCSK et présentons une méthodologie pour l'identifier, en commençant par reconnaître les différents modèles de CSK dans la fabrication et en les alignant sur les concepts de fabrication. L'extraction du MCSK sous une forme utilisable est un défi, c'est pourquoi notre approche structure le MCSK en énoncés NL en utilisant des LLM pour faciliter le raisonnement basé sur des règles, améliorant ainsi les capacités de prise de décision. Le troisième et dernier objectif est de proposer un cadre S-XAI utilisant le RCO et le MCSK pour évaluer si les machines existantes peuvent effectuer des tâches spécifiques et générer des explications NL compréhensibles. Cet objectif a été atteint en intégrant le RCO, qui fournit des capacités opérationnelles telles que la répétabilité et la précision, au MCSK, qui décrit les exigences du processus. En utilisant le raisonnement sémantique basé sur le MCSK, le système S-XAI fournit de manière transparente des explications NL qui détaillent chaque logique et chaque résultat.Dans le cadre du S-XAI, un NN prédit les capacités opérationnelles des robots, tandis que l'IA symbolique incorpore ces prédictions dans un système de raisonnement basé sur le MCSK et fondé sur le RCO.Cette configuration hybride maximise les forces de chaque système d'IA et garantit que les prédictions soutiennent un processus décisionnel transparent. En outre, la S-XAI améliore l'interprétabilité des prédictions du NN grâce à des techniques XAI telles que LIME, SHAP et PDP, clarifiant les prédictions du NN et permettant d'obtenir des informations détaillées pour un meilleur calibrage et une gestion proactive, favorisant ainsi un environnement de fabrication résilient et informéIn Industry 4.0, advanced manufacturing is vital in shaping future factories, enabling enhanced planning, scheduling, and control. The ability to adaptproduction lines swiftly in response to customer demands or unexpected situations is essential to enhance the future of manufacturing. While AI is emerging as a solution, industries still rely on human expertise due to trust issues and a lack of transparency in AI decisions. Explainable AI integrating commonsense knowledge related to manufacturing is crucial for making AI decisions understandable and trustworthy. Within this context, we propose the S-XAI framework, an integrated solution combining machine specifications with MCSK to provide explainable and transparent decision-making. The focus is on providing real-time machine capabilities to ensure precise decision-making while simultaneously explaining the decision-making process to all involved stakeholders. Accordingly, the first objective was formalizing machine specifications, including capabilities, capacities, functions, quality, and process characteristics, focusing on robotics. To do so, we created a Robot Capability ontology formalizing all relevant aspects of machine specifications, such as Capability, Capacity, Function, Quality, and Process Characteristics. On top of this formalization, the RCO allows manufacturing stakeholders to capture robotic capabilities described in specification manuals (advertised capabilities) and compare them with real-world performance (operational capabilities). RCO is based on the Machine Service Description Language, a domain reference ontology created for manufacturing services, and aligned with the Basic Formal Ontology, Industrial Foundry Ontology, Information Artifact Ontology, and Relations Ontology. The second objective was the formalization of MCSK. We introduce MCSK and present a methodology for identifying it, starting with recognizing different CSK patterns in manufacturing and aligning them with manufacturing concepts. Extracting MCSK in a usable form is challenging, so our approach structures MCSK into NL statements utilizing LLMs. to facilitate rule-based reasoning, thereby enhancing decision-making capabilities. The third and final objective is to propose an S-XAI framework utilizing RCO and MCSK to assess if existing machines can perform specific tasks and generate understandable NL explanations. This was achieved by integrating the RCO, which provides operational capabilities like repeatability and precision, with MCSK, which outlines the process requirements. By utilizing MCSK-based semantic reasoning, the S-XAI system seamlessly provides NL explanations that detail each logic and outcome. In the S-XAI framework, an NN predicts the operational capabilities of robots, while symbolic AI incorporates these predictions within an MCSK-based reasoning system grounded in the RCO. This hybrid setup maximizes the strengths of each AI system and ensures that predictions support a transparent decision-making process. Additionally, S-XAI enhances the interpretability of NN predictions through XAI techniques such as LIME, SHAP, and PDP, clarifying NN predictions and enabling detailed insights for better calibration and proactive management, ultimately fostering a resilient and informed manufacturing environment
2
Labeau, Matthieu. "Neural language models : Dealing with large vocabularies." Thesis, Université Paris-Saclay (ComUE), 2018. http://www.theses.fr/2018SACLS313/document.
Full textAbstract:
Le travail présenté dans cette thèse explore les méthodes pratiques utilisées pour faciliter l'entraînement et améliorer les performances des modèles de langues munis de très grands vocabulaires. La principale limite à l'utilisation des modèles de langue neuronaux est leur coût computationnel: il dépend de la taille du vocabulaire avec laquelle il grandit linéairement. La façon la plus aisée de réduire le temps de calcul de ces modèles reste de limiter la taille du vocabulaire, ce qui est loin d'être satisfaisant pour de nombreuses tâches. La plupart des méthodes existantes pour l'entraînement de ces modèles à grand vocabulaire évitent le calcul de la fonction de partition, qui est utilisée pour forcer la distribution de sortie du modèle à être normalisée en une distribution de probabilités. Ici, nous nous concentrons sur les méthodes à base d'échantillonnage, dont le sampling par importance et l'estimation contrastive bruitée. Ces méthodes permettent de calculer facilement une approximation de cette fonction de partition. L'examen des mécanismes de l'estimation contrastive bruitée nous permet de proposer des solutions qui vont considérablement faciliter l'entraînement, ce que nous montrons expérimentalement. Ensuite, nous utilisons la généralisation d'un ensemble d'objectifs basés sur l'échantillonnage comme divergences de Bregman pour expérimenter avec de nouvelles fonctions objectif. Enfin, nous exploitons les informations données par les unités sous-mots pour enrichir les représentations en sortie du modèle. Nous expérimentons avec différentes architectures, sur le Tchèque, et montrons que les représentations basées sur les caractères permettent l'amélioration des résultats, d'autant plus lorsque l'on réduit conjointement l'utilisation des représentations de motsThis work investigates practical methods to ease training and improve performances of neural language models with large vocabularies. The main limitation of neural language models is their expensive computational cost: it depends on the size of the vocabulary, with which it grows linearly. Despite several training tricks, the most straightforward way to limit computation time is to limit the vocabulary size, which is not a satisfactory solution for numerous tasks. Most of the existing methods used to train large-vocabulary language models revolve around avoiding the computation of the partition function, ensuring that output scores are normalized into a probability distribution. Here, we focus on sampling-based approaches, including importance sampling and noise contrastive estimation. These methods allow an approximate computation of the partition function. After examining the mechanism of self-normalization in noise-contrastive estimation, we first propose to improve its efficiency with solutions that are adapted to the inner workings of the method and experimentally show that they considerably ease training. Our second contribution is to expand on a generalization of several sampling based objectives as Bregman divergences, in order to experiment with new objectives. We use Beta divergences to derive a set of objectives from which noise contrastive estimation is a particular case. Finally, we aim at improving performances on full vocabulary language models, by augmenting output words representation with subwords. We experiment on a Czech dataset and show that using character-based representations besides word embeddings for output representations gives better results. We also show that reducing the size of the output look-up table improves results even more
3
Schaeffer, Marion. "Towards efficient Knowledge Graph-based Retrieval Augmented Generation for conversational agents." Electronic Thesis or Diss., Normandie, 2025. http://www.theses.fr/2025NORMIR06.
Full textAbstract:
Les agents conversationnels se sont largement répandus ces dernières années. Aujourd'hui, ils ont dépassé leur objectif initial de simuler une conversation avec un programme informatique et sont désormais des outils précieux pour accéder à l'information et effectuer diverses tâches, allant du service client à l'assistance personnelle. Avec l'essor des modèles génératifs et des grands modèles de langage (LLM), les capacités des agents conversationnels ont été décuplées. Cependant, ils sont désormais sujets à des hallucinations, générant ainsi des informations erronées. Une technique populaire pour limiter le risque d'hallucinations est la génération augmentée par récupération (RAG), qui permet d'injecter des connaissances lors de la génération de texte. Ces connaissances peuvent être extraites de graphes de connaissances (KG), qui sont des représentations structurées et accessibles pour les systèmes informatiques. Ainsi, nous explorons les architectures de KG-RAG pour construire des agents conversationnels de confiance. Nous démontrons l'intérêt de notre approche pour un cas d'usage réel de support citoyen·ne en construisant un agent conversationnel traitant les mesures autour du handicap dans des villes françaises. Nous présentons d'abord un historique des agents conversationnels, en introduisant les méthodes mises en œuvre au fil des années et les techniques d'évaluation. Nous définissons ensuite les KG et les ontologies, et explorons les techniques de construction et d'évaluation. Ne trouvant pas de KG directement exploitable, notre première contribution introduit OLAF : Ontology Learning Applied Framework. Ce système modulaire est conçu pour une construction automatisée et reproductible de KG à partir de textes non structurés. OLAF intègre des techniques linguistiques, statistiques et basées sur des LLM pour générer des ontologies minimales viables sur des domaines spécifiques. Appliqué à des ensembles de données réels, OLAF démontre des performances robustes grâce à des évaluations basées sur des ontologies de référence et des questions de compétence spécifiques à une tâche. Nous détaillons le processus de construction d'un KG sur la thématique du handicap dans une ville française. Nous proposons ensuite une architecture pour les systèmes de KG-RAG afin d'améliorer la recherche d'information en alignant les requêtes des utilisateur·rice·s avec les structures des graphes via la liaison d'entités, les patrons de requêtes et les méthodes de récupération basées sur les LLM. Nous démontrons l'intérêt de notre architecture sur différents cas d'utilisation, que nous évaluons selon des critères tels que la performance, les préférences humaines et l'impact environnemental. Bien que les préférences des utilisateur·rice·s avantagent l'architecture de Text-RAG, l'impact environnemental réduit de l'architecture de KG-RAG souligne son potentiel pour des pratiques d'IA durables. Enfin, nous identifions comme élément clé de l'architecture la partie concernant la recherche d'information. Nous abordons donc cette tâche dans notre architecture en explorant les techniques de vectorisation dans divers contextes, c'est-à-dire en améliorant la liaison d'entités, la recherche des données contextuelles et en fournissant un système de cache. Nous proposons également des mécanismes pour gérer les conversations multi-tours. Ce travail établit un cadre complet pour les systèmes de KG-RAG, combinant la sémantique des KG avec les capacités génératives des LLM pour construire des agents conversationnels précis, spécialisés et durables. Les contributions incluent OLAF pour une construction automatisée de KG, un pipeline de KG-RAG robuste, et des améliorations basées sur des représentations vectorielles pour la précision de la recherche d'information et la qualité des interactions. En répondant aux défis industriels des agents conversationnels, ces travaux posent les bases du déploiement de systèmes de KG-RAG dans des domaines spécialisés et variésConversational agents have become widespread in recent years. Today, they have transcended their initial purpose of simulating a conversation with a computer program and are now valuable tools for accessing information and carrying out various tasks, from customer service to personal assistance. With the rise of text-generative models and Large Language Models (LLMs), the capabilities of conversational agents have increased tenfold. However, they are now subject to hallucinations, producing false information. A popular technique to limit the risk of hallucinations is Retrieval Augmented Generation (RAG), which injects knowledge into a text generation process. Such injected knowledge can be drawn from Knowledge Graphs (KGs), which are structured machine-readable knowledge representations. Therefore, we explore Knowledge Graph-based Retrieval Augmented Generation (KG-RAG) to build trusted conversational agents. We demonstrate our approach on a real-world use case for citizen support by building conversational agents for disability management in cities. We first present a history of conversational agents, introducing the approaches implemented over the years and the evaluation techniques. We then define KGs and ontologies, and explore construction and evaluation techniques. As we could not find a directly exploitable KG, our first contribution introduces the Ontology Learning Applied Framework (OLAF). This modular system is built for automated and repeatable KG construction from unstructured text. OLAF integrates linguistic, statistical, and LLM-based techniques to generate Minimum Viable Ontologies for specific domains. Applied to real-world datasets, OLAF demonstrates robust performance through gold-standard evaluations and task-specific Competency Questions. We detail the construction process for a KG about disability management in a French city. We then propose an architecture for KG-RAG systems to enhance information retrieval by aligning user queries with KG structures through entity linking, graph queries, and LLM-based retrieval approaches. We demonstrate our architecture on different use cases, which we evaluate using criteria such as performance, human preference, and environmental impact. While user preferences advantage Text-RAG, KG-RAG's reduced computational footprint underscores its potential for sustainable AI practices. Finally, we identify the critical part of the architecture as the retriever. Therefore, we tackle the retrieval task in our architecture by exploring embeddings in various contexts, i.e. improving EL, retrieval, and providing a caching system. We also propose mechanisms for handling multi-turn conversations. This work establishes a comprehensive framework for KG-RAG systems, combining the semantic depth of KGs with the generative capabilities of LLMs to deliver accurate, contextual, and sustainable conversational agents. Contributions include OLAF for scalable KG construction, a robust KG-RAG pipeline, and embedding-based enhancements for retrieval and interaction quality. By addressing conversational agents' industrial challenges, such as scalability, retrieval precision, and conversational coherence, this research lays the foundation for deploying KG-RAG systems in diverse and specialised domains
4
Zervakis, Georgios. "Enriching large language models with semantic lexicons and analogies." Electronic Thesis or Diss., Université de Lorraine, 2023. http://www.theses.fr/2023LORR0039.
Full textAbstract:
Les progrès récents de l'apprentissage profond et des réseaux de neurones ont permis d'aborder des tâches complexes de traitement du langage naturel, qui sont appliquées à une pléthore de problèmes réels allant des assistants intelligents dans les appareils mobiles à la prédiction du cancer. Néanmoins, les systèmes modernes basés sur ces approches présentent plusieurs limitations qui peuvent compromettre leurs performances et leur fiabilité, les rendre injustes envers les minorités ou exposer des données personnelles. Nous sommes convaincus que l'intégration de connaissances et de raisonnement symboliques dans le cadre de l'apprentissage profond est une étape nécessaire vers la résolution de ces limitations. Par exemple, les ressources lexicales peuvent enrichir les réseaux de neurones profonds avec des connaissances sémantiques ou syntaxiques, et les règles logiques peuvent fournir des mécanismes d'apprentissage et de raisonnement. Par conséquent, l'objectif de cette thèse est de développer et d'évaluer des moyens d'intégrer différents types de connaissances et de raisonnement symboliques dans un modèle de langage largement utilisé, le Bidirectional Encoder R presentations from Transformers (BERT). Dans un premier temps, nous considérons le retrofitting, une technique simple et populaire pour raffiner les plongements lexicaux de mots grâce à des relations provenant d'un lexique sémantique. Nous présentons deux méthodes inspirées par cette technique pour incorporer ces connaissances dans des plongements contextuels de BERT. Nous évaluons ces méthodes sur trois jeux de données biomédicales pour l'extraction de relations et un jeu de données de critiques de films pour l'analyse des sentiments, et montrons qu'elles n'ont pas d'impact substantiel sur les performances pour ces tâches. En outre, nous effectuons une analyse qualitative afin de mieux comprendre ce résultat négatif. Dans un second temps, nous intégrons le raisonnement analogique à BERT afin d'améliorer ses performances sur la tâche de vérification du sens d'un mot, et de le rendre plus robuste. Pour cela, nous reformulons la vérification du sens d'un mot comme une tâche de détection d'analogie. Nous présentons un modèle hybride qui combine BERT pour encoder les données d'entrée en quadruplets et un classifieur neuronal convolutif pour décider s'ils constituent des analogies valides. Nous testons notre système sur un jeu de données de référence et montrons qu'il peut surpasser les approches existantes. Notre étude empirique montre l'importance de l'encodage d'entrée pour BERT, et comment cette dépendance est atténuée en intégrant les propriétés axiomatiques des analogies lors de l'apprentissage, tout en préservant les performances et en améliorant la robustesseRecent advances in deep learning and neural networks have made it possible to address complex natural language processing tasks, which find application in a plethora of real-world problems ranging from smart assistants in mobile devices to the prediction of cancer. Nonetheless, modern systems based on these frameworks exhibit various limitations that may compromise their performance and trustworthiness, render them unfair towards minorities, or subject them to privacy leakage. It is our belief that integrating symbolic knowledge and reasoning into the deep learning framework is a necessary step towards addressing the aforementioned limitations. For example, lexical resources can enrich deep neural networks with semantic or syntactic knowledge, and logical rules can provide learning and reasoning mechanisms. Therefore, the scope of this thesis is to develop and evaluate ways of integrating different types of symbolic knowledge and reasoning into a widely used language model, Bidirectional Encoder Representations from Transformers (BERT). ln a first stage, we consider retrofitting, a simple and popular technique for refining distributional word embeddings based on relations coming from a semantic lexicon. Inspired by this technique, we present two methods for incorporating this knowledge into BERT contextualized embeddings. We evaluate these methods on three biomedical datasets for relation extraction and one movie review dataset for sentiment analysis, and show that they do not substantially impact the performance for these tasks. Furthermore, we conduct a qualitative analysis to provide further insights on this negative result. ln a second stage, we integrate analogical reasoning with BERT as a means to improve its performance on the target sense verification task, and make it more robust. To do so, we reformulate target sense verification as an analogy detection task. We present a hybrid model that combines BERT to encode the input data into quadruples and a convolutional neural classifier to decide whether they constitute valid analogies. We test our system on a benchmark dataset, and show that it can outperform existing approaches. Our empirical study shows the importance of the input encoding for BERT, and how this dependence gets alleviated by integrating the axiomatic properties of analogies during training, while preserving performance and improving robustness
5
Chadha, Vikrampal. "Simulation of large-scale system-level models." Thesis, This resource online, 1994. http://scholar.lib.vt.edu/theses/available/etd-12162009-020334/.
Full text
6
Bughio, Kulsoom Saima. "IoMT security: A semantic framework for vulnerability detection in remote patient monitoring." Thesis, Edith Cowan University, Research Online, Perth, Western Australia, 2024. https://ro.ecu.edu.au/theses/2841.
Full textAbstract:
The increasing need to safeguard patient data in Internet of Medical Things (IoMT) devices highlights the critical importance of reducing vulnerabilities within these systems. The widespread adoption of IoMT has transformed healthcare by enabling continuous remote patient monitoring (RPM), which enhances patient outcomes and optimizes healthcare delivery. However, the integration of IoMT devices into healthcare systems presents significant security challenges, particularly in protecting sensitive patient data and ensuring the reliability of medical devices. The diversity of data formats used by various vendors in RPM complicates data aggregation and fusion, thereby hindering overall cybersecurity efforts.
This thesis proposes a novel semantic framework for vulnerability detection in RPM settings within the IoMT system. The framework addresses interoperability, heterogeneity, and integration challenges through meaningful data aggregation. The core of this framework is a domain ontology that captures the semantics of concepts and properties related to the primary security aspects of IoT medical devices. This ontology is supported by a comprehensive ruleset and complex queries over aggregated knowledge. Additionally, the implementation integrates medical device data with the National Vulnerability Database (NVD) via an API, enabling real-time detection of vulnerabilities and improving the security of RPM systems.
By capturing the semantics of medical devices and network components, the proposed semantic model facilitates partial automation in detecting network anomalies and vulnerabilities. A logic-based ruleset enhances the system’s robustness and efficiency, while its reasoning capabilities enable the identification of potential vulnerabilities and anomalies in IoMT systems, thereby improving security measures in remote monitoring settings.
The semantic framework also supports knowledge graph visualization and efficient querying through SPARQL. The knowledge graph provides a structured representation of interconnected data and stores Cyber Threat Intelligence (CTI) to enhance data integration, visualization, and semantic enrichment. The query mechanism enables healthcare providers to extract valuable insights from IoMT data, notifying them about new system vulnerabilities or vulnerable medical devices. This demonstrates the impact of vulnerabilities on cybersecurity requirements (Confidentiality, Integrity, and Availability) and facilitates countermeasures based on severity. Consequently, the framework promotes timely decision-making, enhancing the overall efficiency and effectiveness of IoMT systems. The semantic framework is validated through various use cases and existing frameworks, demonstrating its effectiveness and robustness in vulnerability detection within the domain of IoMT security.
7
Hittner, Brian Edward. "Rendering large-scale terrain models and positioning objects in relation to 3D terrain." Thesis, Monterey, Calif. : Springfield, Va. : Naval Postgraduate School ; Available from National Technical Information Service, 2003. http://library.nps.navy.mil/uhtbin/hyperion-image/03Dec%5FHittner.pdf.
Full textAbstract:
Thesis (M.S. in Modeling, Virtual Environments and Simulation)--Naval Postgraduate School, December 2003.Thesis advisor(s): Don Brutzman, Curt Blais. Includes bibliographical references (p. 117-118). Also available online.
8
Kropff, Emilio. "Statistical and dynamical properties of large cortical network models: insights into semantic memory and language." Doctoral thesis, SISSA, 2007. http://hdl.handle.net/20.500.11767/4639.
Full textAbstract:
This thesis introduces several variants to the classical autoassociative memory model in order to capture different characteristics of large cortical networks, using semantic memory as a paradigmatic example in which to apply the results. Chapter 2 is devoted to the development of the sparse Potts model network as a simplification of a multi modular memory performing computations both at the local and the global level. If a network storing p global patterns has N local modules, each one active in S possible ways with a global sparseness a, and if each module is connected to cM other modules, the storage capacity scales like αc ≡ pmax /cM ∝ S 2 /a with logarithmic corrections. Chapter 3 further introduces adaptation and correlations among patterns, as a result of which a latching dynamics appears, consistent in the spontaneous hopping between global attractor states after an initial cue-guided
retrieval, somehow similar to a free association process. The complexity of the latching series depends on the equilibrium between self-excitation of the local networks and global inhibition represented by the parameter U. Finally, Chapter 4 develops a consistent way to store and retrieve correlated patterns, which works as long as any statistical dependence between units can be neglected. The popularity of units must be introduced into the learning rule,
as a result of which a new property of associative memories appears: the robustness of a memory is inverse to the information it conveys. As in some accounts of semantic memory deficits, random damage results in selective impairments, associated to the entropy measure Sf of each memory, since the minimum connectivity required to sustain its retrieval is, in optimal conditions, cM ∝ pSf , and still proportional to pSf but possibly with a larger coefficient in the general case. Present in the entire thesis, but specially in this last Chapter, the conjecture stating that autoassociative memories are limited in the amount of information stored per synapse results consistent with the results.
9
Zhao, Ying, and ying zhao@rmit edu au. "Effective Authorship Attribution in Large Document Collections." RMIT University. Computer Science and Information Technology, 2008. http://adt.lib.rmit.edu.au/adt/public/adt-VIT20080730.162501.
Full textAbstract:
Techniques that can effectively identify authors of texts are of great importance in scenarios such as detecting plagiarism, and identifying a source of information. A range of attribution approaches has been proposed in recent years, but none of these are particularly satisfactory; some of them are ad hoc and most have defects in terms of scalability, effectiveness, and computational cost. Good test collections are critical for evaluation of authorship attribution (AA) techniques. However, there are no standard benchmarks available in this area; it is almost always the case that researchers have their own test collections. Furthermore, collections that have been explored in AA are usually small, and thus whether the existing approaches are reliable or scalable is unclear. We develop several AA collections that are substantially larger than those in literature; machine learning methods are used to establish the value of using such corpora in AA. The results, also used as baseline results in this thesis, show that the developed text collections can be used as standard benchmarks, and are able to clearly distinguish between different approaches. One of the major contributions is that we propose use of the Kullback-Leibler divergence, a measure of how different two distributions are, to identify authors based on elements of writing style. The results show that our approach is at least as effective as, if not always better than, the best existing attribution methods-that is, support vector machines-for two-class AA, and is superior for multi-class AA. Moreover our proposed method has much lower computational cost and is cheaper to train. Style markers are the key elements of style analysis. We explore several approaches to tokenising documents to extract style markers, examining which marker type works the best. We also propose three systems that boost the AA performance by combining evidence from various marker types, motivated from the observation that there is no one type of marker that can satisfy all AA scenarios. To address the scalability of AA, we propose the novel task of authorship search (AS), inspired by document search and intended for large document collections. Our results show that AS is reasonably effective to find documents by a particular author, even within a collection consisting of half a million documents. Beyond search, we also propose the AS-based method to identify authorship. Our method is substantially more scalable than any method published in prior AA research, in terms of the collection size and the number of candidate authors; the discrimination is scaled up to several hundred authors.
10
Pan, Bi-Yu. "Hierarchical test generation for VHDL behavioral models." Thesis, This resource online, 1992. http://scholar.lib.vt.edu/theses/available/etd-09052009-040449/.
Full textBooks on the topic "Large Language Models (LLM)"
1
Amaratunga, Thimira. Understanding Large Language Models. Berkeley, CA: Apress, 2023. http://dx.doi.org/10.1007/979-8-8688-0017-7.
Full text
2
Martra, Pere. Large Language Models Projects. Berkeley, CA: Apress, 2024. http://dx.doi.org/10.1007/979-8-8688-0515-8.
Full text
3
Kucharavy, Andrei, Octave Plancherel, Valentin Mulder, Alain Mermoud, and Vincent Lenders, eds. Large Language Models in Cybersecurity. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7.
Full text
4
Marcondes, Francisco S., Adelino Gala, Renata Magalhães, Fernando Perez de Britto, Dalila Durães, and Paulo Novais. Natural Language Analytics with Generative Large-Language Models. Cham: Springer Nature Switzerland, 2025. https://doi.org/10.1007/978-3-031-76631-2.
Full text
5
Kamath, Uday, Kevin Keenan, Garrett Somers, and Sarah Sorenson. Large Language Models: A Deep Dive. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-65647-7.
Full text
6
Singh, Bhawna. Building Applications with Large Language Models. Berkeley, CA: Apress, 2024. http://dx.doi.org/10.1007/979-8-8688-0569-1.
Full text
7
Grigorov, Dilyan. Introduction to Python and Large Language Models. Berkeley, CA: Apress, 2024. http://dx.doi.org/10.1007/979-8-8688-0540-0.
Full text
8
Mualla, Yazan, Liuwen Yu, Davide Liga, Igor Tchappi, and Réka Markovich, eds. Advances in Explainability, Agents, and Large Language Models. Cham: Springer Nature Switzerland, 2025. https://doi.org/10.1007/978-3-031-89103-8.
Full text
9
Törnberg, Petter. How to Use Large-Language Models for Text Analysis. 1 Oliver’s Yard, 55 City Road, London EC1Y 1SP United Kingdom: SAGE Publications Ltd, 2024. http://dx.doi.org/10.4135/9781529683707.
Full text
10
Jonnagaddala, Jitendra, Hong-Jie Dai, and Ching-Tai Chen, eds. Large Language Models for Automatic Deidentification of Electronic Health Record Notes. Singapore: Springer Nature Singapore, 2025. https://doi.org/10.1007/978-981-97-7966-6.
Full textBook chapters on the topic "Large Language Models (LLM)"
1
Da Silva Gameiro, Henrique. "LLM Detectors." In Large Language Models in Cybersecurity, 197–204. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_22.
Full textAbstract:
AbstractLLM detectors aim at detecting text generated by an LLM. They can be categorized into two main types: specific detectors and general detectors. Specific detectors target a particular type of language or context, such as hate speech or spam. In contrast, general detectors aim to identify a broad range of problematic languages, such as misinformation or propaganda. They typically rely on supervised learning, using large labeled datasets to train the models to recognize patterns in the language. General-purpose detectors have shown bad results, but specific-purpose detectors have shown more promising results. This has to be nuanced due to the broad range of effective attacks, especially the paraphrasing attacks, to which all defense techniques are somewhat vulnerable. There are also many other challenges for developing detectors such as the growing numbers of different LLMs (open source or not) being developed and an effective detector that works with many human languages besides English. Mitigation techniques include storing user conversations with an LLM and watermarking (especially cryptographic).
2
Kucharavy, Andrei. "Overview of Existing LLM Families." In Large Language Models in Cybersecurity, 31–44. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_3.
Full textAbstract:
AbstractWhile the general public discovered Large Language Models (LLMs) with ChatGPT—a generative autoregressive model, they are far from the only models in the LLM family. Various architectures and training regiments optimized for specific usages were designed throughout their development, which were then classified as different LLM families.
3
Würsch, Maxime, Dimitri Percia David, and Alain Mermoud. "Monitoring Emerging Trends in LLM Research." In Large Language Models in Cybersecurity, 153–61. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_17.
Full textAbstract:
AbstractEstablished methodologies for monitoring and forecasting trends in technological development fall short of capturing advancements in Large Language Models (LLMs). This chapter suggests a complementary and alternative approach to mitigate this concern. Traditional indicators, such as search volumes and citation frequencies, are demonstrated to inadequately reflect the rapid evolution of LLM-related technologies due to biases, semantic drifts, and inherent lags in data documentation. Our presented methodology analyzes the proximity of technological terms related to LLMs, leveraging the OpenAlex and arXiv databases, and focuses on extracting nouns from scientific papers to provide a nuanced portrayal of advancements in LLM technologies. The approach aims to counteract the inherent lags in data, accommodate semantic drift, and distinctly differentiate between various topics, offering both retrospective and prospective insights in their analytical purview. The insights derived underline the need for refined, robust, adaptable, and precise forecasting models as LLMs intersect with domains like cyber defense. At the same time, they are considering the limitations of singular ontologies and integrating advanced anticipatory measures for a nuanced understanding of evolving LLM technologies.
4
Meier, Raphael. "LLM-Aided Social Media Influence Operations." In Large Language Models in Cybersecurity, 105–12. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_11.
Full textAbstract:
AbstractSocial media platforms enable largely unrestricted many-to-many communication. In times of crisis, they offer a space for collective sense-making and give rise to new social phenomena (e.g., open-source investigations). However, they also serve as a tool for threat actors to conduct Cyber-enabled Social Influence Operations (CeSIOs) to shape public opinion and interfere in decision-making processes. CeSIOs employ sock puppet accounts to engage authentic users in online communication, exert influence, and subvert online discourse. Large Language Models (LLMs) may further enhance the deceptive properties of sock puppet accounts. Recent LLMs can generate targeted and persuasive text, which is, for the most part, indistinguishable from human-written content—ideal features for covert influence. This article reviews recent developments at the intersection of LLMs and influence operations, summarizes LLMs’ salience, and explores the potential impact of LLM-instrumented sock puppet accounts for CeSIOs. Finally, mitigation measures for the near future are highlighted.
5
Majumdar, Subhabrata. "Standards for LLM Security." In Large Language Models in Cybersecurity, 225–31. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_25.
Full textAbstract:
AbstractThe National Institute of Standards and Technology (NIST) is a recognized authority on computer security that publishes guidelines and standards for a broad range of technologies, including artificial intelligence (AI). The guidelines include the requirement for LLM decision-making transparency, explainability, testing, and validation to guarantee model reliability and security. Moreover, the NIST has also created standards for cryptography, a critical element of many LLM-based applications, such as secure communication and data encryption. The cryptography standards help ensure that LLM-based applications are secure and resilient against attacks by malicious entities. NIST standards can provide a practical framework for secure and ethical LLM-based application development and deployment. By adhering to these standards, developers and organizations can increase the confidence that their LLM-based applications are dependable, trustworthy, and resistant to attacks.
6
Schillaci, Zachary. "LLM Adoption Trends and Associated Risks." In Large Language Models in Cybersecurity, 121–28. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_13.
Full textAbstract:
AbstractThe emergence of Large Language Models (LLMs) is expected to impact the job market significantly, accelerating automation trends and posing a risk to traditionally creative-oriented jobs. LLMs can automate tasks in various fields, including design, journalism, and creative writing. Companies and public institutions can leverage generative models to enhance productivity and reduce workforce requirements through machine-assisted workflows and natural language interactions. While technical skills like programming may become less important in certain roles, generative models are unlikely to fully replace programmers due to the need for expertise in code validation and niche development. The enterprise landscape of LLMs comprises providers (organizations training proprietary models), integrators (technology companies fine-tuning LLMs for specific applications), and users (companies and individuals adopting LLM-powered solutions). The applications of the models include conversational search, customer service chatbots, content creation, personalized marketing, data analysis, and basic workflow automation. The regulatory landscape is rapidly evolving, with key considerations including copyright, data security, and liability. Government involvement and informed expertise are recommended to guide governance and decision-making processes in this domain.
7
Martra, Pere. "Creating and Publishing Your Own LLM." In Large Language Models Projects, 297–318. Berkeley, CA: Apress, 2024. http://dx.doi.org/10.1007/979-8-8688-0515-8_7.
Full text
8
Li, Yixuan, Julian Parsert, and Elizabeth Polgreen. "Guiding Enumerative Program Synthesis with Large Language Models." In Computer Aided Verification, 280–301. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-65630-9_15.
Full textAbstract:
AbstractPre-trained Large Language Models (LLMs) are beginning to dominate the discourse around automatic code generation with natural language specifications. In contrast, the best-performing synthesizers in the domain of formal synthesis with precise logical specifications are still based on enumerative algorithms. In this paper, we evaluate the abilities of LLMs to solve formal synthesis benchmarks by carefully crafting a library of prompts for the domain. When one-shot synthesis fails, we propose a novel enumerative synthesis algorithm, which integrates calls to an LLM into a weighted probabilistic search. This allows the synthesizer to provide the LLM with information about the progress of the enumerator, and the LLM to provide the enumerator with syntactic guidance in an iterative loop. We evaluate our techniques on benchmarks from the Syntax-Guided Synthesis (SyGuS) competition. We find that GPT-3.5 as a stand-alone tool for formal synthesis is easily outperformed by state-of-the-art formal synthesis algorithms, but our approach integrating the LLM into an enumerative synthesis algorithm shows significant performance gains over both the LLM and the enumerative synthesizer alone and the winning SyGuS competition tool.
9
Apaydin, Kaan, and Yorck Zisgen. "Local Large Language Models for Business Process Modeling." In Lecture Notes in Business Information Processing, 605–9. Cham: Springer Nature Switzerland, 2025. https://doi.org/10.1007/978-3-031-82225-4_44.
Full textAbstract:
Abstract Large language models (LLMs) are capable of efficiently understanding natural language by processing large volumes of text data. Natural language is also used in process descriptions, thus LLMs appear to be a suitable candidate to significantly improve business process modeling. Although plenty of third-party LLMs exist, they raise the risk of privacy disclosure, untrustworthiness, and generalizability of the results. This paper proposes a pipeline to use a local and fine-tuned LLM that expects a textual process description as input and finally generates a visual process tree representation. We instantiate our pipeline with Llama3 8B and fine-tune the LLM with a training set of 120 self-generated examples. Initial evaluation results of our LLM-based approach for automated business process modeling promise usefulness of the approach in terms of process model quality while preserving data privacy.
10
Vogelsang, Terry. "LLM Controls Execution Flow Hijacking." In Large Language Models in Cybersecurity, 99–104. Cham: Springer Nature Switzerland, 2024. http://dx.doi.org/10.1007/978-3-031-54827-7_10.
Full textAbstract:
AbstractLLMs can be vulnerable to prompt injection attacks. Similar to how code injections can alter the behavior of a given program, malicious prompt injection can influence the execution flow of a specific business logic. This is due to their reliance on user-provided text for controlling execution flow. In the context of interactive systems, this poses significant business and cybersecurity risks. Mitigations such as prohibiting the use of LLMs in critical systems, developing prompt and resulting API calls verification tools, implementing security by designing good practices, and enhancing incident logging and alerting mechanisms can be considered to reduce the novel attack surface presented by LLMs.
Conference papers on the topic "Large Language Models (LLM)"
1
Pasupuleti, Rajesh, Ravi Vadapalli, Christopher Mader, and Norris Timothy. "Popular LLM-Large Language Models in Enterprise Applications." In 2024 2nd International Conference on Foundation and Large Language Models (FLLM), 125–31. IEEE, 2024. https://doi.org/10.1109/fllm63129.2024.10852443.
Full text
2
Fernando, Saruni, Robert Kunzelmann, Daniela Sánchez Lopera, Jad Al Halabi, and Wolfgang Ecker. "Boosting Productivity of Hardware Documentation Using Large Language Models." In 2024 IEEE LLM Aided Design Workshop (LAD), 1. IEEE, 2024. http://dx.doi.org/10.1109/lad62341.2024.10691698.
Full text
3
Vijayaraghavan, Prashanth, Luyao Shi, Ehsan Degan, and Xin Zhang. "CircuitSynth: Leveraging Large Language Models for Circuit Topology Synthesis." In 2024 IEEE LLM Aided Design Workshop (LAD), 1–6. IEEE, 2024. http://dx.doi.org/10.1109/lad62341.2024.10691716.
Full text
4
Coppolillo, Erica, Francesco Calimeri, Giuseppe Manco, Simona Perri, and Francesco Ricca. "LLASP: Fine-tuning Large Language Models for Answer Set Programming." In 21st International Conference on Principles of Knowledge Representation and Reasoning {KR-2023}, 834–44. California: International Joint Conferences on Artificial Intelligence Organization, 2024. http://dx.doi.org/10.24963/kr.2024/78.
Full textAbstract:
Recently, Large Language Models (LLMs) have showcased their potential in various natural language processing tasks, including code generation. However, while significant progress has been made in adapting LLMs to generate code for several imperative programming languages and tasks, there remains a notable gap in their application to declarative formalisms, such as Answer Set Programming (ASP). In this paper, we move a step towards exploring the capabilities of LLMs for ASP code generation. First, we perform a systematic evaluation of several state-of-the-art LLMs. Despite their power in terms of number of parameters, training data and computational resources, empirical results demonstrate inadequate performances in generating correct ASP programs. Therefore, we propose LLASP, a fine-tuned lightweight model specifically trained to encode fundamental ASP program patterns. To this aim, we create an ad-hoc dataset covering a wide variety of fundamental problem specifications that can be encoded in ASP. Our experiments demonstrate that the quality of ASP programs generated by LLASP is remarkable. This holds true not only when compared to the non-fine-tuned counterpart but also when compared to the majority of eager LLM candidates, particularly from a semantic perspective. All the code and data used to perform the experiments are publicly available: https://github.com/EricaCoppolillo/LLASP.
5
Jignasu, Anushrut, Kelly Marshall, Baskar Ganapathysubramanian, Aditya Balu, Chinmay Hegde, and Adarsh Krishnamurthy. "Evaluating Large Language Models for G-Code Debugging, Manipulation, and Comprehension." In 2024 IEEE LLM Aided Design Workshop (LAD), 1–5. IEEE, 2024. http://dx.doi.org/10.1109/lad62341.2024.10691700.
Full text
6
Xu, Haocheng, Haotian Hu, and Sitao Huang. "Optimizing High-Level Synthesis Designs with Retrieval-Augmented Large Language Models." In 2024 IEEE LLM Aided Design Workshop (LAD), 1–5. IEEE, 2024. http://dx.doi.org/10.1109/lad62341.2024.10691855.
Full text
7
Petrovic, Nenad, Krzysztof Lebioda, Vahid Zolfaghari, André Schamschurko, Sven Kirchner, Nils Purschke, Fengjunjie Pan, and Alois Knoll. "LLM-Driven Testing for Autonomous Driving Scenarios." In 2024 2nd International Conference on Foundation and Large Language Models (FLLM), 173–78. IEEE, 2024. https://doi.org/10.1109/fllm63129.2024.10852505.
Full text
8
Ru Teh, Jocelyn Shuang, Eng Keong Koay, Shin Wei Lim, Kuan Heng Lee, Mee Sim Lai, Meng Siong Lee, and Yuan Kuok Nee. "Adaptive Composite Accuracy Scoring for Domainspecific LLM Evaluation." In 2024 2nd International Conference on Foundation and Large Language Models (FLLM), 272–79. IEEE, 2024. https://doi.org/10.1109/fllm63129.2024.10852484.
Full text
9
Zolfaghari, Vahid, Nenad Petrovic, Fengjunjie Pan, Krzysztof Lebioda, and Alois Knoll. "Adopting RAG for LLM-Aided Future Vehicle Design." In 2024 2nd International Conference on Foundation and Large Language Models (FLLM), 437–42. IEEE, 2024. https://doi.org/10.1109/fllm63129.2024.10852467.
Full text
10
Loukil, Faiza, Sarah Cadereau, Hervé Verjus, Mattéo Galfre, Kavé Salamatian, David Telisson, Quentin Kembellec, and Olivier Le Van. "LLM-centric pipeline for information extraction from invoices." In 2024 2nd International Conference on Foundation and Large Language Models (FLLM), 569–75. IEEE, 2024. https://doi.org/10.1109/fllm63129.2024.10852504.
Full textReports on the topic "Large Language Models (LLM)"
1
Zhang, Hao. Large Language Model (LLM) Monthly Report (2024 Apr). ResearchHub Technologies, Inc., May 2024. http://dx.doi.org/10.55277/researchhub.0ps6xenm.
Full text
2
Azuara Herrera, Oliver, Laura Ripani, and Eric Torres Ramirez. AI and the Increase of Productivity and Labor Inequality in Latin America: Potential Impact of Large Language Models on Latin American Workforce. Inter-American Development Bank, September 2024. http://dx.doi.org/10.18235/0013152.
Full textAbstract:
We assess the potential effect of large language models (LLMs) on the labor markets of Chile, Mexico, and Peru using the methodology of Eloundou et al. (2023). This approach involves detailed guidelines (rubrics) for each job to assess whether access to LMM software would reduce the time required for workers to complete their daily tasks. Adapting this methodology to the Latin American context necessitated developing a comprehensive crosswalk between the Occupational Information Network (O*NET) and regional occupational classifications, SINCO-2011 and ISCO-2008. When we use this adaptation, the theoretical average task exposure of occupations under these classifications is 32% and 31% for each classification. Recognizing the unique characteristics of each country's labor market, we refined these ratings to reflect better each nation's capacity to adopt and effectively implement new technologies. After these adjustments, the task exposure for SINCO-2011 drops to 27% and for ISCO-2008 to 23%. These adjusted exposure ratings provide a more accurate depiction of the real-world implications of LLM integration in the Latin American context. According to this methodology, the LLM-powered exposure using GPT-4 estimates suggests that the percentage of jobs with task exposure exceeding 10% is 74% in Mexico, 76% in Chile, and 76% in Peru. When we raise the exposure threshold to 40% or more, the proportion of affected occupations significantly decreases to 9% in Mexico, 20% in Chile, and 6% in Peru. The exposure is close to zero after this threshold. In other words, the exposure would only affect less than half of the total labor force in these countries. Further analysis of exposure by socioeconomic conditions indicates higher exposure among women, individuals with higher education, formal employees, and higher-income groups. This suggests a potential increase in labor inequality in the region due to adopting this technology. Our findings highlight the need for targeted policy interventions and adaptive strategies to ensure that the transition to an AI-enhanced labor market benefits all socio-economic groups and minimizes disruptions.
3
Rosenblat, Sruly, Tim O'Reilly, and Ilan Strauss. Beyond Public Access in LLM Pre-Training Data: Non-public book content in OpenAI’s Models. AI Disclosures Project, Social Science Research Council, April 2025. https://doi.org/10.35650/aidp.4111.d.2025.
Full textAbstract:
Using a legally obtained dataset of 34 copyrighted O’Reilly Media books, we apply the DE-COP membership inference attack method to investigate whether OpenAI’s large language models were trained on copyrighted content without consent. Our AUROC scores show that GPT-4o, OpenAI’s more recent and capable model, demonstrates strong recognition of paywalled O’Reilly book content (AUROC = 82%), compared to OpenAI’s earlier model GPT-3.5 Turbo. In contrast, GPT-3.5 Turbo shows greater relative recognition of publicly accessible O’Reilly book samples. GPT-4o Mini, as a much smaller model, shows no knowledge of public or non-public O’Reilly Media content when tested (AUROC ≈ 0.5). Testing multiple models, with the same cutoff date, helps us account for potential language shifts over time that might bias our findings. These results highlight the urgent need for increased corporate transparency regarding pre-training data sources as a means to develop formal licensing frameworks for AI content training.
4
Alonso-Robisco, Andres, and Jose Manuel Carbo. Analysis of CBDC Narrative OF Central Banks using Large Language Models. Madrid: Banco de España, August 2023. http://dx.doi.org/10.53479/33412.
Full textAbstract:
Central banks are increasingly using verbal communication for policymaking, focusing not only on traditional monetary policy, but also on a broad set of topics. One such topic is central bank digital currency (CBDC), which is attracting attention from the international community. The complex nature of this project means that it must be carefully designed to avoid unintended consequences, such as financial instability. We propose the use of different Natural Language Processing (NLP) techniques to better understand central banks’ stance towards CBDC, analyzing a set of central bank discourses from 2016 to 2022. We do this using traditional techniques, such as dictionary-based methods, and two large language models (LLMs), namely Bert and ChatGPT, concluding that LLMs better reflect the stance identified by human experts. In particular, we observe that ChatGPT exhibits a higher degree of alignment because it can capture subtler information than BERT. Our study suggests that LLMs are an effective tool to improve sentiment measurements for policy-specific texts, though they are not infallible and may be subject to new risks, like higher sensitivity to the length of texts, and prompt engineering.
5
Marra de Artiñano, Ignacio, Franco Riottini Depetris, and Christian Volpe Martincus. Automatic Product Classification in International Trade: Machine Learning and Large Language Models. Inter-American Development Bank, July 2023. http://dx.doi.org/10.18235/0005012.
Full textAbstract:
Accurately classifying products is essential in international trade. Virtually all countries categorize products into tariff lines using the Harmonized System (HS) nomenclature for both statistical and duty collection purposes. In this paper, we apply and assess several different algorithms to automatically classify products based on text descriptions. To do so, we use agricultural product descriptions from several public agencies, including customs authorities and the United States Department of Agriculture (USDA). We find that while traditional machine learning (ML) models tend to perform well within the dataset in which they were trained, their precision drops dramatically when implemented outside of it. In contrast, large language models (LLMs) such as GPT 3.5 show a consistently good performance across all datasets, with accuracy rates ranging between 60% and 90% depending on HS aggregation levels. Our analysis highlights the valuable role that artificial intelligence (AI) can play in facilitating product classification at scale and, more generally, in enhancing the categorization of unstructured data.
6
Moreno, Ángel Iván, and Teresa Caminero. Assessing the data challenges of climate-related disclosures in european banks. A text mining study. Madrid: Banco de España, September 2023. http://dx.doi.org/10.53479/33752.
Full textAbstract:
The Intergovernmental Panel on Climate Change (IPCC) estimates that global net-zero should be achieved by 2050. To this end, many private firms are pledging to reach net-zero emissions by 2050. The Climate Data Steering Committee (CDSC) is working on an initiative to create a global central digital repository of climate disclosures, which aims to address the current data challenges. This paper assesses the progress within European financial institutions towards overcoming the data challenges outlined by the CDSC. Using a text-mining approach, coupled with the application of commercial Large Language Models (LLM) for context verification, we calculate a Greenhouse Gas Disclosure Index (GHGDI), by analysing 23 highly granular disclosures in the ESG reports between 2019 and 2021 of most of the significant banks under the ECB’s direct supervision. This index is then compared with the CDP score. The results indicate a moderate correlation between institutions not reporting to CDP upon request and a low GHGDI. Institutions with a high CDP score do not necessarily correlate with a high GHGDI.
7
Harrison, Stephen. Wikipedia’s Governance Challenge: Policies and Guardrails for New Generative AI Technologies. Balsillie School of International Affairs, November 2024. https://doi.org/10.51644/bcs002.
Full textAbstract:
The Wikipedia editing community is struggling with the emergence of new generative artificial intelligence (AI) technologies such as ChatGPT and other large language models (LLMs). Should the volunteer editors who create and maintain Wikipedia’s articles be permitted to use new generative AI tools, or should they be prohibited because of the high risk of introducing misinformation into the popular internet encyclopedia?
8
Maerz, Seraphine. Using AI for Text Analysis in R. Instats Inc., 2024. http://dx.doi.org/10.61700/ti5uexui5ilrd1663.
Full textAbstract:
This one-day workshop provides a basic introduction to using artificial intelligence for text analysis. Tailored for researchers across a variety of social and health science fields, participants will gain practical skills in building text corpuses, topic modeling, and using ChatGPT, Copilot, and other Large Language Models (LLMs) in R, while addressing ethical considerations and validation techniques for AI-driven research.
9
Korinek, Anton, and Jai Vipra. Concentrating Intelligence: Scaling and Market Structure in Artificial Intelligence. Institute for New Economic Thinking Working Paper Series, October 2024. http://dx.doi.org/10.36687/inetwp228.
Full textAbstract:
This paper examines the evolving structure and competition dynamics of the rapidly growing market for foundation models, focusing on large language models (LLMs). We describe the technological characteristics that shape the industry and have given rise to fierce competition among the leading players. The paper analyzes the cost structure of foundation models, emphasizing the importance of key inputs such as computational resources, data, and talent, and identifies significant economies of scale and scope that may create a tendency towards greater market concentration in the future. We explore two concerns for competition, the risk of market tipping and the implications of vertical integration, and use our analysis to inform policy remedies to maintain a competitive landscape.
10
Bastidas Ripalda, Rafaela, Stephen Hansen, John Leon-Diaz, and Yabra Muvdi. Tracking the Reform Process from Newspaper Data. Inter-American Development Bank, March 2025. https://doi.org/10.18235/0013467.
Full textAbstract:
Countries often undertake structural reforms to boost productivity, reduce inequality, and improve public sector efficiency, yet there is limited data on the reform process to evaluate these efforts. We examine the extent to which large language models (LLMs) applied to media articles can fill this evidence gap, using a case study from Colombia. We find that suitably prompted LLMs produce output that aligns with human reading. The output reveals extensive coverage of the debate and discussion that precedes reform adoption or rejection, and can be used to track the evolution of specific reforms. We conclude that our methodology, once expanded, could produce a uniquely valuable dataset.10.18235/0013467