A Survey of Large Language Models: Evolution, Architectures, Adaptation, Benchmarking, Applications, Challenges, and Societal Implications

مسح نماذج اللغة الكبيرة: التطور والبنى والتكييف والقياس والتطبيقات والتحديات والآثار المجتمعية

👤 Seyed Mahmoud Sajjadi Mohammadabadi, Burak Cem Kara, Can Eyupoglu, Can Uzay, Mehmet Serkan Tosun, Oktay Karakus 📄 Electronics, 2025 · 14(18), 3580 · 31 pages 🔗 10.3390/electronics14183580 ✓ CC BY 4.0

الملخص

يقدم هذا المسح مراجعة متعمقة لنماذج اللغة الكبيرة (LLMs)، مسلطًا الضوء على التحول النموذجي الكبير الذي تمثله في الذكاء الاصطناعي. وهدفنا هو توحيد أحدث التطورات في تصميم LLM وتدريبها وتكييفها وتقييمها وتطبيقها لخدمة الباحثين والممارسين. ولتحقيق ذلك نتتبع تطور نماذج اللغة ونصف المناهج الأساسية بما فيها الضبط الدقيق الفعال بالمعاملات (PEFT). وتشمل المنهجية مسحًا شاملًا لتطبيقات LLM الواقعية عبر القطاعات العلمية والهندسية والصحية والإبداعية، إلى جانب مراجعة المعايير الحالية.

تشير نتائجنا إلى أن تكاليف التدريب والاستدلال المرتفعة تشكل بنية السوق وتثير مخاوف اقتصادية وعمالية، مع التأكيد على الحاجة المستمرة للإشراف البشري في التقييم. ومن الاتجاهات الرئيسية تطوير بنى متعددة الوسائط موحدة تعالج مدخلات بيانات متنوعة، وظهور أنظمة وكيلة تُظهر سلوكيات معقدة مثل استخدام الأدوات والتخطيط. ونحدد مشكلات مفتوحة حرجة مثل قابلية الكشف وتلوث البيانات والتعميم وتنوع المعايير. ونخلص إلى أن تجاوز هذه التحديات التقنية والاقتصادية والاجتماعية المعقدة يتطلب تقدمًا تعاونيًا في التكييف والتقييم والبنية التحتية والحوكمة.

1. المقدمة

تمثل نماذج اللغة الكبيرة (LLMs) تحولًا نموذجيًا في الذكاء الاصطناعي يتجاوز بكثير معالجة اللغة الطبيعية (NLP) التقليدية. وتُظهر النماذج التأسيسية مثل سلسلة GPT من OpenAI [1-3] وPaLM من Google [4] وLLaMA من Meta [5] قدرات جديدة على فهم اللغة البشرية والاستدلال وتوليدها عبر مجالات تمتد من هندسة البرمجيات والاكتشاف العلمي إلى التعليم والفنون الإبداعية. وتُبرز هذه القدرات في جوهرها قوة التعلم الذاتي الإشراف (SSL) على نطاق واسع، حيث تكتسب النماذج معرفة عميقة قابلة للتعميم من بيتابايتات النصوص غير الموسومة والبيانات متعددة الوسائط.

ويتيرة الابتكار استثنائية، مع ظهور سريع لبنى وتدريبات وتطبيقات جديدة — لكن هذا النمو المتفجر خلق جسدًا معرفيًا مجزأً يصعب على الباحثين والممارسين وصناع السياسات الإحاطة به. ومع تعمق اندماج LLM في البنية المجتمعية، أصبح الفهم الواضح لتقنياتها وقدراتها وقيودها ضروريًا. ينظم هذا المسح المجال عبر تتبع التطور، ومقارنة البنى، وشرح التكييف، ومراجعة المعايير، ورسم التطبيقات القطاعية، وتحليل الآثار الاقتصادية والمجتمعية، بما فيها صعود الوكلاء المستقلين القادرين على استخدام الأدوات والتخطيط.

2. تطور نمذجة اللغة

تطورت نمذجة اللغة عبر أربع حقب. اعتمدت الأنظمة القائمة على القواعد (قبل التسعينيات)، مثل ELIZA والقوالب النحوية [8،9]، على قواعد لغوية مصنوعة يدويًا هشة خارج النطاقات الضيقة. وأدخلت النماذج الإحصائية (التسعينيات–الألفينات) احتمالات n-gram المقدّرة من المدونات، محققة تعميمًا قائمًا على البيانات لكنها عانت من التشتت ونافذة سياق قصيرة ثابتة [10-12]. واستبدلت النماذج العصبية التسلسلية (الألفينات–العشرينيات) العدّ بالتمثيلات الموزعة: نموذج Bengio الاحتمالي العصبي [13] وشبكات Elman [14] وLSTM [15] وتضمينات word2vec [17] التي التقطت اعتماديات أطول وتشابهًا دلاليًا.

وأزالت حقبة Transformer (أواخر العشرينيات–الحاضر) التكرار تمامًا: فبنية Vaswani وآخرين القائمة على الانتباه فقط [20] مكّنت التدريب المتوازي على نطاق غير مسبوق، مولّدة GPT (تدريب مسبق أحادي الاتجاه [1،2]) وBERT (نمذجة مقنّعة ثنائية الاتجاه [21]) وخلفاءهما. ثم أظهرت قوانين التوسع أن الأداء يتحسن بشكل متوقع مع المعاملات والبيانات والحوسبة، مشعلة السباق نحو نماذج بمئات المليارات من المعاملات وقدرات ناشئة للتعلم من أمثلة قليلة [3،27].

3. بنى النماذج

تُدرَّب نماذج فاكّ التشفير فقط (عائلة GPT وLLaMA وPaLM [1-5]) بالتنبؤ السببي للرمز التالي وتهيمن على التطبيقات التوليدية؛ وتستخدم نماذج المشفّر فقط (BERT وRoBERTa [21]) النمذجة المقنّعة ثنائية الاتجاه وتتفوق في التصنيف والاستخلاص؛ وتجمع نماذج التسلسل-إلى-التسلسل (T5 وBART) بين مشفّر ثنائي الاتجاه وفاكّ انحداري ذاتي للترجمة والتلخيص. وتوسّع النماذج متعددة الوسائط بنية Transformer إلى الصور والصوت والمدخلات المهيكلة — مثل المحاذاة التباينية بأسلوب CLIP والانتباه المتقاطع المبوّب بأسلوب Flamingo — ممكنة نماذج موحدة مثل GPT-4V وGemini التي تستدل على النصوص والصور معًا.

وتوسّع بنى خليط الخبراء (MoE) السعة دون حوسبة متناسبة عبر توجيه كل رمز إلى مجموعة فرعية صغيرة من شبكات الخبراء بدالة بوّابة متعلمة، كما في Switch Transformer وMixtral؛ فتخفض عمليات FLOPs النشطة لكل رمز جذريًا لكنها تضيف تعقيد التوجيه وخسائر موازنة الحمل وبصمات ذاكرة كبيرة. وعبر جميع العائلات، يلاحظ المسح اتجاه تقارب: بنى متمحورة حول فاكّ التشفير، متعددة الوسائط، مفعّلة بتشتت، تُدرَّب بنوافذ سياق أطول باستمرار.

4. التدريب والتكييف

يستهلك التدريب المسبق مدونات بحجم الويب بأهداف سببية أو مقنّعة، يليه الضبط الدقيق على بيانات موسومة، وهندسة المطالب والتعلم في السياق اللذان يوجّهان النماذج المجمدة بتعليمات مصممة وأمثلة قليلة، وضبط التعليمات على عروض مهام متنوعة، والتعلم المعزز من التغذية البشرية الراجعة (RLHF) الذي يوائم المخرجات مع التفضيلات عبر نموذج مكافأة متعلم وتحسين السياسة [28-34]. وتقايض كل مرحلة العمومية بالمواءمة: فـ RLHF يحسّن الفائدة والسلامة كثيرًا لكنه قد يقلل تنوع المخرجات وينطوي على مخاطر اختراق المكافأة.

يكيّف الضبط الدقيق الفعال بالمعاملات (PEFT) النماذج العملاقة بجزء ضئيل من التكلفة [35]. تحقن الطرق القائمة على المحوّلات وحدات عنق زجاجة صغيرة (إسقاط سفلي، ولا خطية، وإسقاط علوي) بين طبقات Transformer المجمدة وتُدرَّب هي فقط. وتتعلم طرق المطالب (المطالب الناعمة) متجهات مهام مستمرة تُلحق بالمدخلات. وتجمّد طرق إعادة المعاملات مثل LoRA الأوزان الأساسية وتتعلم مصفوفات تحديث منخفضة الرتبة، مضاهية جودة الضبط الكامل في مهام كثيرة مع تدريب أقل من 1% من المعاملات — مما يجعل التكييف المجالي ممكنًا على وحدة GPU واحدة.

5. القياس والتقييم

في القياس، يختبر MMLU المعرفة والاستدلال عبر 57 مادة أكاديمية ومهنية بإعداد الأمثلة القليلة، مقيسًا التعميم خارج توزيعات التدريب [40]؛ ويساهم BIG-bench بأكثر من 200 مهمة مجتمعية تستهدف القدرات الناشئة مثل الفكاهة والاستدلال الحسابي والحكم الأخلاقي؛ ويركّز SuperGLUE على الفهم اللغوي العميق بمهام الاستلزام وإحالة الضمائر الصعبة؛ ويقيّم HELM النماذج شموليًا عبر الدقة والمتانة والعدالة والكفاءة وأبعاد أخرى بدل الدرجة الواحدة.

وفي مقاييس التقييم، تُعد ROUGE عائلة متركزة على الاستدعاء للتلخيص تُبلغ الدقة والاستدعاء وF1 على تداخل n-gram مع المراجع البشرية [44]، بينما BLEU نظيرها المتركز على الدقة للترجمة، معاقِبًا المخرجات القصيرة بعقوبة إيجاز. ويعتمد كلاهما على التداخل السطحي ويرتبطان ارتباطًا ناقصًا بالحكم البشري، مما يحفز الحكام القائمين على النماذج — مع تأكيد المسح على أن تكاليف الاستدلال المرتفعة وتلوث البيانات المستمر (تسرب بنود القياس إلى بيانات التدريب) يُبقيان الإشراف البشري لا غنى عنه في التقييم.

6. تطبيقات LLM

يرسم المسح تسعة قطاعات. ففي هندسة البرمجيات والتصميم تدعم LLM إكمال الشيفرة وتوليد الاختبارات وإصلاح البرامج؛ وفي الرعاية الصحية وعلوم الحياة تساعد في التوثيق السريري وروبوتات الفرز وتنقيب أدبيات البروتينات تحت إشراف صارم؛ وفي المالية والمصارف تقود تحليل المشاعر وصياغة التقارير وكشف أنماط الاحتيال [63]؛ وفي التصنيع وسلاسل الإمداد تدعم أدلة الصيانة التنبؤية وتخطيط اللوجستيات؛ وفي البحث العلمي تسرّع تركيب الأدبيات وتوليد الفرضيات وأتمتة شيفرة المختبرات.

وفي التعليم والتدريب المؤسسي تمكّن حوارات التدريس وتوليد التقييمات؛ وفي الصناعات الإبداعية والمحتوى تشارك في كتابة النصوص التسويقية والسيناريوهات والسرديات؛ وفي القطاعات القانونية والتنظيمية تلخص السوابق القضائية وترصد مخاطر الامتثال؛ ويتسلسل الوكلاء المستقلون (Auto-GPT وأنظمة LangChain [78،79]) المهام الفرعية ويستدعون واجهات API الخارجية وأنظمة الملفات والمتصفحات لتنفيذ عمليات رقمية معقدة — وهو المثال الأبرز في المسح على تطور LLM من روبوتات محادثة إلى فاعلين.

7. الآثار الاقتصادية لتطوير LLM ونشرها

ترتكز اقتصاديات LLM على تكاليف دورة حياة هائلة: فالتدريب المسبق لنموذج تأسيسي يتطلب آلاف وحدات GPU أو TPU المتطورة العاملة لأسابيع أو أشهر — نفقات رأسمالية مقدمة تصاعدت من ميزانيات حقبة GPT-2 إلى جولات بمئات ملايين الدولارات — بينما يضيف تقديم الاستدلال على نطاق واسع تكاليف حوسبة وعرض نطاق ذاكرة وطاقة مستمرة (الجدول 7). وتركّز هذه النفقات التطوير في حفنة منظمات غنية بالموارد، دافعة توحيد السوق حول مزودي API وبائعي السحابة، والتسويق عبر الاشتراكات والتسعير حسب الاستخدام وخدمات الضبط للمؤسسات.

وتواجه أسواق العمل ضغطًا مزدوجًا: فأتمتة الصياغة والترميز والتحليل والدعم ترفع الإنتاجية لكنها تزيح العمل المعرفي الروتيني وتوسّع الفجوة بين الشركات المتبنية للذكاء الاصطناعي والمتخلفة [92]. ويجادل المسح بأن تكاليف التدريب والاستدلال المرتفعة تشكل بنية السوق فعلًا، وأن الاستجابات السياسية — الوصول الحاسوبي ورقابة المنافسة وإعادة تأهيل القوى العاملة — يجب أن ترافق التقدم التقني.

8. الاتجاهات الحديثة والقضايا المفتوحة

تهيمن خمسة اتجاهات: بنى متعددة الوسائط موحدة تعالج النصوص والصور والصوت والفيديو أصليًا؛ وكشف المحتوى المولّد عبر العلامات المائية (إشارات سرية قابلة للكشف إحصائيًا [157،158]) وتتبع المصدر والمصنفات في سباق تسلح مع المولدات؛ وLLM الوكيلة باستدلال معزز بالأدوات وتخطيط وتنفيذ متعدد الخطوات؛ ودفعات قابلية التوسع والكفاءة (الانتباه المتشتت والتكميم والتقطير وسياقات أطول مثل ادعاءات RULER [180،181])؛ والأخلاقيات والتنظيم والأثر المجتمعي عبر التحيز والخصوصية وحقوق النشر.

وتستمر ست مشكلات مفتوحة (الشكل 7): المفاضلة بين قابلية الكشف وسهولة الاستخدام؛ وتلوث بيانات المعايير؛ والتعميم خارج توزيعات التدريب؛ وتنوع المعايير وتشبعها؛ ومواءمة السلوك الوكيلي مع القصد البشري؛ وحوكمة القدرات مزدوجة الاستخدام. ويؤكد المسح أن حلها يتطلب تقدمًا مشتركًا في منهجية التقييم وشفافية بيانات التدريب والبنى الفعالة والأطر التنظيمية — فلا يكفي حل واحد.

9. الاستنتاجات

يوحّد هذا المسح مشهد LLM من الأصول القائمة على القواعد عبر Transformer وتشتت MoE وتكييف PEFT ومواءمة RLHF، إلى المعايير (MMLU وBIG-bench وHELM) والمقاييس (ROUGE وBLEU) وقطاعات التطبيق التسعة واقتصاديات التدريب والنشر. وخلاصته المحورية أن التوسع جلب تغيرًا نوعيًا — التعلم من أمثلة قليلة واتباع التعليمات واستخدام الأدوات — لكنه جلب أيضًا تكاليف هيكلية: أسواق مركزة واضطراب عمالي وهشاشة تقييم من التلوث والتشبع.

ويخلص المسح إلى أن تجاوز هذه التحديات التقنية والاقتصادية والاجتماعية المتشابكة يتطلب تقدمًا تعاونيًا في مناهج التكييف وعلم التقييم والبنية التحتية الفعالة والحوكمة. فـ LLM تصبح بنية مجتمعية؛ وتوجيهها نحو المنفعة العامة يتطلب من الباحثين والصناعة وصناع السياسات التقدم معًا بدل تحسين محور واحد وحده.

Abstract

This survey provides an in-depth review of large language models (LLMs), highlighting the significant paradigm shift they represent in artificial intelligence. Our purpose is to consolidate state-of-the-art advances in LLM design, training, adaptation, evaluation, and application for both researchers and practitioners. To accomplish this, we trace the evolution of language models and describe core approaches, including parameter-efficient fine-tuning (PEFT). The methodology involves a thorough survey of real-world LLM applications across the scientific, engineering, healthcare, and creative sectors, coupled with a review of current benchmarks.

Our findings indicate that high training and inference costs are shaping market structures, raising economic and labor concerns, while also underscoring a persistent need for human oversight in assessment. Key trends include the development of unified multimodal architectures capable of processing varied data inputs and the emergence of agentic systems that exhibit complex behaviors such as tool use and planning. We identify critical open problems, such as detectability, data contamination, generalization, and benchmark diversity. Ultimately, we conclude that overcoming these complex technical, economic, and social challenges necessitates collaborative advancements in adaptation, evaluation, infrastructure, and governance.

1. Introduction

Large language models (LLMs) represent a paradigm shift in artificial intelligence, extending far beyond traditional natural language processing (NLP). Foundational models such as OpenAI's GPT series [1–3], Google's PaLM [4], and Meta's LLaMA [5] show novel abilities to understand, reason, and generate human language across fields from software engineering and scientific discovery to education and creative arts. At their core, these capabilities highlight the power of self-supervised learning (SSL) at scale, where models acquire deep, generalizable knowledge from petabytes of unlabeled text and multimodal data.

The pace of innovation is extraordinary, with rapid introductions of new architectures, training methodologies, and applications — but this explosive growth has created a fragmented body of knowledge that challenges researchers, practitioners, and policymakers to navigate. As LLMs integrate deeper into societal infrastructure, a clear understanding of their technologies, capabilities, and limitations is essential. This survey organizes the field by tracing evolution, comparing architectures, explaining adaptation, reviewing benchmarks, mapping sector-wise applications, and analyzing economic and societal implications, including the rise of autonomous agents with tool use and planning.

2. Evolution of Language Modeling

Language modeling evolved through four eras. Rule-based systems (pre-1990s), such as ELIZA and template grammars [8,9], relied on handcrafted linguistic rules that were brittle outside narrow domains. Statistical models (1990s–2000s) introduced n-gram probabilities estimated from corpora, bringing data-driven generalization but suffering sparsity and a fixed short context window [10–12]. Sequential neural models (2000s–2020s) replaced counting with distributed representations: Bengio's neural probabilistic LM [13], Elman networks [14], LSTMs [15], and word2vec embeddings [17] captured longer dependencies and semantic similarity.

The Transformer era (late 2010s–present) removed recurrence entirely: Vaswani et al.'s attention-only architecture [20] enabled parallel training at unprecedented scale, giving rise to GPT (decoder-only pre-training [1,2]), BERT (bidirectional masked modeling [21]), and their successors. Scaling laws then showed that performance improves predictably with parameters, data, and compute, triggering the race toward hundred-billion-parameter models and emergent few-shot abilities [3,27].

3. Model Architectures

Decoder-only models (GPT family, LLaMA, PaLM [1–5]) train with causal next-token prediction and dominate generative applications; encoder-only models (BERT, RoBERTa [21]) use bidirectional masked modeling and excel at classification and extraction; sequence-to-sequence models (T5, BART) pair a bidirectional encoder with an autoregressive decoder for translation and summarization. Multimodal models extend Transformers to vision, audio, and structured inputs — e.g. CLIP-style contrastive alignment and Flamingo-style gated cross-attention — enabling unified models such as GPT-4V and Gemini that reason jointly over text and images.

Mixture-of-Experts (MoE) architectures scale capacity without proportional compute by routing each token to a small subset of expert feed-forward networks with a learned gating function, as in Switch Transformer and Mixtral; they cut active FLOPs per token dramatically but add routing complexity, load-balancing losses, and large memory footprints. Across all families, the survey notes a convergence trend: decoder-centric, multimodal, sparsely activated architectures trained with ever-longer context windows.

4. Training and Adaptation

Pre-training consumes web-scale corpora with causal or masked objectives, followed by fine-tuning on labeled data, prompt engineering and in-context learning that steer frozen models with crafted instructions and few-shot examples, instruction tuning on diverse task demonstrations, and reinforcement learning from human feedback (RLHF), which aligns outputs with preferences via a learned reward model and policy optimization [28–34]. Each stage trades generality for alignment: RLHF greatly improves helpfulness and safety but can reduce output diversity and risks reward hacking.

Parameter-efficient fine-tuning (PEFT) adapts giant models at a fraction of the cost [35]. Adapter-based methods inject small bottleneck modules (down-projection, non-linearity, up-projection) between frozen Transformer layers and train only them. Prompt-based (soft-prompt) methods learn continuous task vectors prepended to inputs. Reparameterization methods such as LoRA freeze the base weights and learn low-rank update matrices, matching full fine-tuning quality on many tasks while training under 1% of parameters — making domain adaptation feasible on single GPUs.

5. Benchmarking and Evaluation

For benchmarking, MMLU tests knowledge and reasoning across 57 academic and professional subjects in a few-shot setup, measuring generalization beyond training distributions [40]; BIG-bench contributes 200+ community-designed tasks targeting emergent abilities such as humor, arithmetic reasoning, and moral judgment; SuperGLUE stresses deep language understanding with difficult entailment and coreference tasks; and HELM evaluates models holistically across accuracy, robustness, fairness, efficiency, and other dimensions rather than a single score.

For evaluation metrics, ROUGE is a recall-oriented family for summarization that reports precision, recall, and F1 over n-gram overlap with human references [44], while BLEU is its precision-oriented counterpart for translation, penalizing short outputs with a brevity penalty. Both rely on surface overlap and correlate imperfectly with human judgment, motivating model-based judges — though the survey stresses that high inference costs and persistent data contamination (benchmark items leaking into training data) keep human oversight indispensable in assessment.

6. Applications of LLMs

The survey maps nine sectors. In software engineering and design, LLMs power code completion, test generation, and program repair; in healthcare and life sciences they assist clinical documentation, triage chatbots, and protein-literature mining under strict oversight; in finance and banking they drive sentiment analysis, report drafting, and fraud-pattern detection [63]; in manufacturing and supply chains they support predictive maintenance manuals and logistics planning; and in scientific research they accelerate literature synthesis, hypothesis generation, and lab-code automation.

In education and corporate training they enable tutoring dialogues and assessment generation; in creative and content industries they co-write marketing copy, scripts, and game narratives; in legal and regulatory sectors they summarize case law and flag compliance risks; and autonomous AI agents (Auto-GPT, LangChain-based systems [78,79]) chain subtasks and call external APIs, file systems, and browsers to execute complex digital operations — the survey's prime example of LLMs evolving from chatbots into actors.

7. Economic Implications of LLM Development and Deployment

LLM economics are anchored by immense lifecycle costs: pre-training a foundation model demands thousands of high-end GPUs or TPUs running for weeks or months — a front-loaded capital outlay that has escalated from GPT-2-era budgets to hundred-million-dollar runs — while serving inference at scale adds ongoing compute, memory-bandwidth, and energy costs (Table 7). These expenditures concentrate development in a handful of well-resourced organizations, driving market consolidation around API providers and cloud vendors, and commercialization via subscriptions, usage-based pricing, and enterprise fine-tuning services.

Labor markets face dual pressure: automation of drafting, coding, analysis, and support roles raises productivity but displaces routine cognitive work and widens inequality between AI-adopting firms and laggards [92]. The survey argues that high training and inference costs are actively shaping market structure and that policy responses — compute access, competition oversight, and workforce retraining — must accompany technical progress.

8. Recent Trends and Open Issues

Five trends dominate: unified multimodal architectures that natively process text, image, audio, and video; detection of generated content via watermarking (secret statistically detectable signals [157,158]), provenance tracking, and classifiers in an arms race with generators; agentic LLMs with tool-augmented reasoning, planning, and multi-step execution; scalability and efficiency pushes (sparse attention, quantization, distillation, and longer contexts such as RULER-evaluated claims [180,181]); and ethics, regulation, and societal impact spanning bias, privacy, and copyright.

Six open problems persist (Figure 7): the detectability-vs-usability trade-off; data contamination of benchmarks; generalization beyond training distributions; benchmark diversity and saturation; alignment of agentic behavior with human intent; and governance of dual-use capabilities. Solving them, the survey stresses, requires joint progress on evaluation methodology, training-data transparency, efficient architectures, and regulatory frameworks — no single fix suffices.

9. Conclusions

This survey consolidates the LLM landscape from rule-based origins through Transformers, MoE sparsity, PEFT adaptation, and RLHF alignment, to benchmarks (MMLU, BIG-bench, HELM), metrics (ROUGE, BLEU), nine application sectors, and the economics of training and deployment. Its central finding is that scale brought qualitative change — few-shot learning, instruction following, tool use — but also structural costs: concentrated markets, labor disruption, and evaluation fragility from contamination and saturation.

The survey concludes that overcoming these intertwined technical, economic, and social challenges demands collaborative advances in adaptation methods, evaluation science, efficient infrastructure, and governance. LLMs are becoming societal infrastructure; steering them toward broad benefit requires researchers, industry, and policymakers to progress together rather than optimize any single axis alone.

References

  1. [1] Radford, A.; Narasimhan, K.; Salimans, T.; Sutskever, I. Improving Language Understanding by Generative Pre-Training. OpenAI Technical Report 2018. Available online: https://cdn.openai.com/research-covers/language-unsupervised/language_ understanding_paper.pdf (accessed on 20 June 2025)
  2. [2] Radford, A.; Wu, J.; Child, R.; Luan, D.; Amodei, D.; Sutskever, I. Language models are unsupervised multitask learners. OpenAI Blog 2019, 1, 9.
  3. [3] Brown, T.; Mann, B.; Ryder, N.; Subbiah, M.; Kaplan, J.D.; Dhariwal, P.; Neelakantan, A.; Shyam, P.; Sastry, G.; Askell, A.; et al. Language models are few-shot learners. Adv. Neural Inf. Process. Syst. 2020, 33, 1877–1901.
  4. [4] Chowdhery, A.; Narang, S.; Devlin, J.; Bosma, M.; Mishra, G.; Roberts, A.; Barham, P.; Chung, H.W.; Sutton, C.; Gehrmann, S.; et al. Palm: Scaling language modeling with pathways. J. Mach. Learn. Res. 2023, 24, 1–113.
  5. [5] Touvron, H.; Lavril, T.; Izacard, G.; Martinet, X.; Lachaux, M.A.; Lacroix, T.; Rozière, B.; Goyal, N.; Hambro, E.; Azhar, F.; et al. Llama: Open and efficient foundation language models. arXiv 2023 arXiv:2302.13971.
  6. [6] Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A.N.; Kaiser, Ł.; Polosukhin, I. Attention is all you need. In Proceedings of the 31st Conference on Neural Information Processing Systems 2017, Long Beach, CA, USA, 4–9 December 2017; Advances in Neural Information Processing Systems; Volume 30.
  7. [7] Weizenbaum, J. ELIZA—A computer program for the study of natural language communication between man and machine. Commun. ACM 1966, 9, 36–45.
  8. [8] Colby, K.M.; Weber, S.; Hilf, F.D. Artificial paranoia. Artif. Intell. 1971, 2, 1–25.
  9. [9] Wallace, R.S. The Anatomy of ALICE; Springer: Berlin/Heidelberg, Germany, 2009.
  10. [10] Winograd, T. Procedures as a Representation for Data in a Computer Program for Understanding natural Language; MIT Press: Cambridge, MA, USA, 1971.
  11. [11] Jelinek, F. Statistical Methods for Speech Recognition; MIT Press: Cambridge, MA, USA, 1998.
  12. [12] Rabiner, L.R. A tutorial on hidden Markov models and selected applications in speech recognition. Proc. IEEE 2002, 77, 257–286.
  13. [13] Lafferty, J.; McCallum, A.; Pereira, F. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings of the ICML, Williamstown, MA, USA, 28 June–1 July 2001; Volume 1, p. 3.
  14. [14] Elman, J.L. Finding structure in time. Cogn. Sci. 1990, 14, 179–211.
  15. [15] Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780.
  16. [16] Cho, K.; Van Merriënboer, B.; Gulcehre, C.; Bahdanau, D.; Bougares, F.; Schwenk, H.; Bengio, Y. Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv 2014, arXiv:1406.1078.
  17. [17] Mikolov, T.; Chen, K.; Corrado, G.; Dean, J. Efficient estimation of word representations in vector space. arXiv 2013, arXiv:1301.3781.
  18. [18] Pennington, J.; Socher, R.; Manning, C.D. Glove: Global vectors for word representation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), Doha, Qatar, 25–29 October 2014; pp. 1532–1543.
  19. [19] Devlin, J.; Chang, M.W.; Lee, K.; Toutanova, K. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Minneapolis, MN, USA, 2–7 June 2019; Volume 1, pp. 4171–4186.
  20. [20] Liu, A.; Feng, B.; Xue, B.; Wang, B.; Wu, B.; Lu, C.; Zhao, C.; Deng, C.; Zhang, C.; Ruan, C.; et al. Deepseek-v3 technical report. arXiv 2024, arXiv:2412.19437.
  21. [21] Raffel, C.; Shazeer, N.; Roberts, A.; Lee, K.; Narang, S.; Matena, M.; Zhou, Y.; Li, W.; Liu, P.J. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res. 2020, 21, 1–67.
  22. [22] Liu, Y.; Ott, M.; Goyal, N.; Du, J.; Joshi, M.; Chen, D.; Levy, O.; Lewis, M.; Zettlemoyer, L.; Stoyanov, V. Roberta: A robustly optimized bert pretraining approach. arXiv 2019, arXiv:1907.11692.
  23. [23] Lan, Z.; Chen, M.; Goodman, S.; Gimpel, K.; Sharma, P.; Soricut, R. Albert: A lite bert for self-supervised learning of language representations. arXiv 2019, arXiv:1909.11942.
  24. [24] Anil, R.; Dai, A.M.; Firat, O.; Johnson, M.; Lepikhin, D.; Passos, A.; Shakeri, S.; Taropa, E.; Bailey, P.; Chen, Z.; et al. Palm 2 technical report. arXiv 2023, arXiv:2305.10403.
  25. [25] Lewis, M.; Liu, Y.; Goyal, N.; Ghazvininejad, M.; Mohamed, A.; Levy, O.; Stoyanov, V.; Zettlemoyer, L. Bart: Denoising sequence- to-sequence pre-training for natural language generation, translation, and comprehension. arXiv 2019, arXiv:1910.13461.
  26. [26] Lu, H.; Liu, W.; Zhang, B.; Wang, B.; Dong, K.; Liu, B.; Sun, J.; Ren, T.; Li, Z.; Yang, H.; et al. Deepseek-vl: Towards real-world vision-language understanding. arXiv 2024, arXiv:2403.05525.
  27. [27] Hurst, A.; Lerer, A.; Goucher, A.P.; Perelman, A.; Ramesh, A.; Clark, A.; Ostrow, A.; Welihinda, A.; Hayes, A.; Radford, A.; et al. Gpt-4o system card. arXiv 2024, arXiv:2410.21276.
  28. [28] Shazeer, N.; Mirhoseini, A.; Maziarz, K.; Davis, A.; Le, Q.; Hinton, G.; Dean, J. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. arXiv 2017, arXiv:1701.06538.
  29. [29] Jiang, A.Q.; Sablayrolles, A.; Roux, A.; Mensch, A.; Savary, B.; Bamford, C.; Chaplot, D.S.; Casas, D.d.l.; Hanna, E.B.; Bressand, F.; et al. Mixtral of experts. arXiv 2024, arXiv:2401.04088.
  30. [30] Team, G.; Georgiev, P.; Lei, V.I.; Burnell, R.; Bai, L.; Gulati, A.; Tanzer, G.; Vincent, D.; Pan, Z.; Wang, S.; et al. Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context. arXiv 2024, arXiv:2403.05530.
  31. [31] Fedus, W.; Zoph, B.; Shazeer, N. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. J. Mach. Learn. Res. 2022, 23, 1–39.
  32. [32] Wei, J.; Bosma, M.; Zhao, V.Y.; Guu, K.; Yu, A.W.; Lester, B.; Du, N.; Dai, A.M.; Le, Q.V. Finetuned language models are zero-shot learners. arXiv 2021, arXiv:2109.01652.
  33. [33] Sanh, V.; Webson, A.; Raffel, C.; Bach, S.H.; Sutawika, L.; Alyafeai, Z.; Chaffin, A.; Stiegler, A.; Scao, T.L.; Raja, A.; et al. Multitask prompted training enables zero-shot task generalization. arXiv 2021, arXiv:2110.08207.
  34. [34] Ouyang, L.; Wu, J.; Jiang, X.; Almeida, D.; Wainwright, C.; Mishkin, P.; Zhang, C.; Agarwal, S.; Slama, K.; Ray, A.; et al. Training language models to follow instructions with human feedback. Adv. Neural Inf. Process. Syst. 2022, 35, 27730–27744.
  35. [35] Houlsby, N.; Giurgiu, A.; Jastrzebski, S.; Morrone, B.; De Laroussilhe, Q.; Gesmundo, A.; Attariyan, M.; Gelly, S. Parameter- efficient transfer learning for NLP. In Proceedings of the International Conference on Machine Learning, PMLR, Long Beach, CA, USA, 9–15 June 2019; pp. 2790–2799.
  36. [36] Lester, B.; Al-Rfou, R.; Constant, N. The power of scale for parameter-efficient prompt tuning. arXiv 2021, arXiv:2104.08691.
  37. [37] Li, X.L.; Liang, P. Prefix-tuning: Optimizing continuous prompts for generation. arXiv 2021, arXiv:2101.00190.
  38. [38] Hu, E.J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; Chen, W. Lora: Low-rank adaptation of large language models. ICLR 2022, 1, 3.
  39. [39] Dettmers, T.; Pagnoni, A.; Holtzman, A.; Zettlemoyer, L. Qlora: Efficient finetuning of quantized llms. Adv. Neural Inf. Process. Syst. 2023, 36, 10088–10115.
  40. [40] Hendrycks, D.; Burns, C.; Basart, S.; Zou, A.; Mazeika, M.; Song, D.; Steinhardt, J. Measuring massive multitask language understanding. arXiv 2020, arXiv:2009.03300.
  41. [41] Srivastava, A.; Rastogi, A.; Rao, A.; Shoeb, A.A.M.; Abid, A.; Fisch, A.; Brown, A.R.; Santoro, A.; Gupta, A.; Garriga-Alonso, A.; et al. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv 2022, arXiv:2206.04615.
  42. [42] Wang, A.; Pruksachatkun, Y.; Nangia, N.; Singh, A.; Michael, J.; Hill, F.; Levy, O.; Bowman, S. Superglue: A stickier benchmark for general-purpose language understanding systems. Adv. Neural Inf. Process. Syst. 2019, 32. https://doi.org/10.48550/arXiv.19
  43. [43] Liang, P.; Bommasani, R.; Lee, T.; Tsipras, D.; Soylu, D.; Yasunaga, M.; Zhang, Y.; Narayanan, D.; Wu, Y.; Kumar, A.; et al. Holistic evaluation of language models. arXiv 2022, arXiv:2211.09110.
  44. [44] Lin, C.Y. Rouge: A package for automatic evaluation of summaries. In Proceedings of the Text Summarization Branches Out, Barcelona, Spain, 25–26 July 2004; pp. 74–81.
  45. [45] Papineni, K.; Roukos, S.; Ward, T.; Zhu, W.J. Bleu: A method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, Philadelphia, PA, USA, 7–12 July 2002; pp. 311–318.
  46. [46] Eger, S.; Cao, Y.; D’Souza, J.; Geiger, A.; Greisinger, C.; Gross, S.; Hou, Y.; Krenn, B.; Lauscher, A.; Li, Y.; et al. Transforming Science with Large Language Models: A Survey on AI-assisted Scientific Discovery, Experimentation, Content Generation, and Evaluation. arXiv 2025, arXiv:2502.05151.
  47. [47] Whitfield, S.; Hofmann, M.A. Elicit: AI literature review research assistant. Public Serv. Q. 2023, 19, 201–207.
  48. [48] Ramos, M.C.; Collison, C.J.; White, A.D. A review of large language models and autonomous agents in chemistry. Chem. Sci. 2025, 16, 2514–2572.
  49. [49] Frincu, I. In Search of the Perfect Prompt. Master’s Thesis, Aalto University, Espoo, Finland, 2023.
  50. [50] Borhani, M.; Soofiani, N.M.; Ebrahimi, E.; Asadollah, S. First Report on Growth and Reproduction of Turcinoemacheilus bahaii (Esmaeili, Sayyadzadeh, Özulug, Geiger and Freyhof, 2014), in Zayandeh Roud River, Iran. Austin Environ. Sci. 2017, 2, 1014. https://doi.org/10.26420/austinenvironsci.2017.1014.
  51. [51] Nazi, Z.A.; Peng, W. Large language models in healthcare and medical domain: A review. Informatics 2024, 11, 57.
  52. [52] Mohammadabadi, S.M.S.; Peikani, M.B. Identification and classification of rheumatoid arthritis using artificial intelligence and machine learning. In Diagnosing Musculoskeletal Conditions using Artifical Intelligence and Machine Learning to Aid Interpretation of Clinical Imaging; Elsevier: Amsterdam, The Netherlands, 2025; pp. 123–145.
  53. [53] Zheng, Y.; Koh, H.Y.; Yang, M.; Li, L.; May, L.T.; Webb, G.I.; Pan, S.; Church, G. Large language models in drug discovery and development: From disease mechanisms to clinical trials. arXiv 2024, arXiv:2409.04481.
  54. [54] Mohammadabadi, S.M.S.; Seyedkhamoushi, F.; Mostafavi, M.; Peikani, M.B. Examination of AI’s role in Diagnosis, Treatment, and Patient care. In Transforming Gender-Based Healthcare with AI and Machine Learning; CRC Press: Boca Raton, FL, USA, 2024; pp. 221–238.
  55. [55] Huynh, N.; Lin, B. Large Language Models for Code Generation: A Comprehensive Survey of Challenges, Techniques, Evaluation, and Applications. arXiv 2025, arXiv:2503.01245.
  56. [56] Wong, M.F.; Guo, S.; Hang, C.N.; Ho, S.W.; Tan, C.W. Natural language generation and understanding of big code for AI-assisted programming: A review. Entropy 2023, 25, 888.
  57. [57] Liu, B.; Jiang, Y.; Zhang, Y.; Niu, N.; Li, G.; Liu, H. An Empirical Study on the Potential of LLMs in Automated Software Refactoring. arXiv 2024, arXiv:2411.04444.
  58. [58] Zhong, R.; Du, X.; Kai, S.; Tang, Z.; Xu, S.; Zhen, H.L.; Hao, J.; Xu, Q.; Yuan, M.; Yan, J. Llm4eda: Emerging progress in large language models for electronic design automation. arXiv 2023, arXiv:2401.12224.
  59. [59] Mohammadabadi, S.M.S.; Entezami, M.; Moghaddam, A.K.; Orangian, M.; Nejadshamsi, S. Generative artificial intelligence for distributed learning to enhance smart grid communication. Int. J. Intell. Netw. 2024, 5, 267–274.
  60. [60] Mohammadabadi, S.M.S.; Liu, Y.; Canafe, A.; Yang, L. Towards distributed learning of pmu data: A federated learning based event classification approach. In Proceedings of the 2023 IEEE Power & Energy Society General Meeting (PESGM), Orlando, FL, USA, 16–20 July 2023; IEEE: New York, NY, USA, 2023; pp. 1–5.
  61. [61] Li, Y.; Wang, S.; Ding, H.; Chen, H. Large language models in finance: A survey. In Proceedings of the Fourth ACM International Conference on AI in Finance, Brooklyn, NY, USA, 27–29 November 2023; pp. 374–382.
  62. [62] Peddinti, S.R.; Katragadda, S.R.; Pandey, B.K.; Tanikonda, A. Utilizing large language models for advanced service management: Potential applications and operational challenges. J. Sci. Technol. 2023, 4, doi:10.2139/ssrn.5119925.
  63. [63] Lopez-Lira, A.; Kwon, J.; Yoon, S.; Sohn, J.y.; Choi, C. Bridging language models and financial analysis. arXiv 2025, arXiv:2503.22693.
  64. [64] Sriram, A. Comparative Forecasting in Retail Supply Chains Using Machine Learning and Large Language Models. Master’s Thesis, State University of New York at Binghamton, Binghamton, NY, USA, 2025.
  65. [65] Brundage, M.P.; Sharp, M.; Pavel, R. Qualifying evaluations from human operators: Integrating sensor data with natural language logs. PHME Soc. Eur. Conf. 2021, 6, 9.
  66. [66] Bakas, N.P.; Papadaki, M.; Vagianou, E.; Christou, I.; Chatzichristofis, S.A. Integrating llms in higher education, through interactive problem solving and tutoring: Algorithmic approach and use cases. In Proceedings of the European, Mediterranean, and Middle Eastern Conference on Information Systems, Dubai, United Arab Emirates, 11–12 December 2023; Springer: Cham, Switzerland, 2023; pp. 291–307.
  67. [67] McHugh, B.; Myers, D.; Patel, A. AI Co-Counsel: An Attorney’s Guide to Using Artificial Intelligence in the Practice of Law Symposium. Akron Law Rev. 2024, 57, 3.
  68. [68] Khikmatillaeva, M. Beyond Chatbots: How specialized AI tools are reducing legal workloads. FARS Int. J. Educ. Soc. Sci. Humanit. 2025, 13, 133–154.
  69. [69] Cheng, S. When Journalism meets AI: Risk or opportunity? Digit. Gov. Res. Pract. 2025, 6, 1–12.
  70. [70] Dhariwal, P.; Jun, H.; Payne, C.; Kim, J.W.; Radford, A.; Sutskever, I. Jukebox: A generative model for music. arXiv 2020, arXiv:2005.00341.
  71. [71] Topirceanu, A.; Barina, G.; Udrescu, M. Musenet: Collaboration in the music artists industry. In Proceedings of the 2014 European Network Intelligence Conference, Wroclaw, Poland, 29–30 September 2014; IEEE: New York, NY, USA, 2014; pp. 89–94.
  72. [72] Marcus, G.; Davis, E.; Aaronson, S. A very preliminary analysis of DALL-E 2. arXiv 2022, arXiv:2204.13807.
  73. [73] Lu, Y. Artificial Intelligence Applied On Today‘s Urban and Architectural Conceptual Design-A Competition Case Study. Ph.D. Thesis, Politecnico di Torino, Torino, Italy, 2025.
  74. [74] Shetye, S. An evaluation of khanmigo, a generative ai tool, as a computer-assisted language learning app. Stud. Appl. Linguist. TESOL 2024, 24. https://doi.org/10.52214/salt.v24i1.12869.
  75. [75] Vega, J.; Rodriguez, M.; Check, E.; Moran, H.; Loo, L. Duolingo evolution: From automation to artificial intelligence. In Proceedings of the IEEE Colombian Conference on Applications of Computational Intelligence, Pamplona, Colombia, 17–19 July 2024; Springer: Berlin/Heidelberg, Germany, 2025; pp. 54–71.
  76. [76] Calamas, D. Student and Instructor Feedback on an AI-Assisted Grading Tool; American Society for Engineering Education: Washing- ton, DC, USA, 2024.
  77. [77] Hidalgo-Reyes, J.; Alvarez, J.; Guevara-Chavez, L.; Cruz-Netro, Z.G. Gradescope as a Tool to Improve Assessment and Feedback in Engineering. In Proceedings of the 2025 Institute for the Future of Education Conference (IFE), Monterrey, Mexico, 28–30 January 2025; IEEE: New York, NY, USA, 2025; pp. 1–7.
  78. [78] Yang, H.; Yue, S.; He, Y. Auto-gpt for online decision making: Benchmarks and additional opinions. arXiv 2023, arXiv:2306.02224.
  79. [79] O’BRIEN, P.D.; Wiegand, M.E. Agent based process management: Applying intelligent agents to workflow. Knowl. Eng. Rev. 1998, 13, 161–174.
  80. [80] Van Veen, D.; Van Uden, C.; Blankemeier, L.; Delbrouck, J.-B.; Aali, A.; Bluethgen, C.; Pareek, A.; Polacin, M.; Reis, E.P.; Seehofnerová, A.; et al. Clinical Text Summarization: Adapting Large Language Models Can Outperform Human Experts. Res. Sq. 2023. https://doi.org/10.21203/rs.3.rs-3483777/v1.
  81. [81] Rao, A.; Kim, J.; Kamineni, M.; Pang, M.; Lie, W.; Dreyer, K.J.; Succi, M.D. Evaluating GPT as an adjunct for radiologic decision making: GPT-4 versus GPT-3.5 in a breast imaging pilot. J. Am. Coll. Radiol. 2023, 20, 990–997.
  82. [82] Singhal, K.; Azizi, S.; Tu, T.; Mahdavi, S.S.; Wei, J.; Chung, H.W.; Scales, N.; Tanwani, A.; Cole-Lewis, H.; Pfohl, S.; et al. Large language models encode clinical knowledge. Nature 2023, 620, 172–180.
  83. [83] Benary, M.; Wang, X.D.; Schmidt, M.; Soll, D.; Hilfenhaus, G.; Nassir, M.; Sigler, C.; Knödler, M.; Keller, U.; Beule, D.; et al. Leveraging large language models for decision support in personalized oncology. JAMA Netw. Open 2023, 6, e2343689–e2343689.
  84. [84] Clusmann, J.; Kolbinger, F.R.; Muti, H.S.; Carrero, Z.I.; Eckardt, J.N.; Laleh, N.G.; Löffler, C.M.L.; Schwarzkopf, S.C.; Unger, M.; Veldhuizen, G.P.; et al. The future landscape of large language models in medicine. Commun. Med. 2023, 3, 141.
  85. [85] Wu, S.; Irsoy, O.; Lu, S.; Dabravolski, V.; Dredze, M.; Gehrmann, S.; Kambadur, P.; Rosenberg, D.; Mann, G. Bloomberggpt: A large language model for finance. arXiv 2023, arXiv:2303.17564.
  86. [86] Völker, T.; Pfister, J.; Koopmann, T.; Hotho, A. From Chat to Publication Management: Organizing your related work using BibSonomy & LLMs. In Proceedings of the 2024 Conference on Human Information Interaction and Retrieval, Sheffield UK, 10–14 March 2024; pp. 386–390.
  87. [87] Rane, N.; Choudhary, S.; Rane, J. Gemini versus ChatGPT: Applications, performance, architecture, capabilities, and implementa- tion. J. Appl. Artif. Intell. 2024, 5, 69–93.
  88. [88] Wang, W.; Gu, L.; Zhang, L.; Luo, Y.; Dai, Y.; Shen, C.; Xie, L.; Lin, B.; He, X.; Ye, J. SciPIP: An LLM-based Scientific Paper Idea Proposer. arXiv 2024, arXiv:2410.23166.
  89. [89] Sırt, M.; Eyüpo˘glu, C. A user-friendly and explainable framework for redesigning AutoML processes with large lan- guage models. In Proceedings of the 33rd Signal Processing and Communications Applications Conference (SIU), Sile, Istanbul, Turkiye, 25–28 June 2025; pp. 1–4. https://doi.org/10.1109/SIU66497.2025.11112150.
  90. [90] Xia, H.; Yang, Z.; Zhao, Y.; Wang, Y.; Li, J.; Tracy, R.; Zhu, Z.; Wang, Y.F.; Chen, H.; Shen, W. Language and multimodal models in sports: a survey of datasets and applications. arXiv 2024, arXiv:2406.12252.
  91. [91] Rahimian, P.; Flisar, J.; Sumpter, D. Automated explanation of machine learning models of footballing actions in words. J. Sport. Anal. 2025, 11, 22150218251353089.
  92. [92] Cottier, B.; Rahman, R.; Fattorini, L.; Maslej, N.; Besiroglu, T.; Owen, D. The rising costs of training frontier AI models. arXiv 2024, arXiv:2405.21015.
  93. [93] Liu, Y.; He, H.; Han, T.; Zhang, X.; Liu, M.; Tian, J.; Zhang, Y.; Wang, J.; Gao, X.; Zhong, T.; et al. Understanding llms: A comprehensive overview from training to inference. Neurocomputing 2025, 620, 129190.
  94. [94] Gale, T.; Elsen, E.; Hooker, S. Do Neural Networks Really Need to Be So Big? 2020. MIT-IBM Watson AI Lab Blog. Available online: https://mitibmwatsonailab.mit.edu/research/blog/do-neural-networks-really-need-to-be-so-big/ (accessed on 1 August 2025).
  95. [95] Buchholz, K. The Extreme Cost of Training AI Models, 2024. Available online: https://www.forbes.com/sites/ katharinabuchholz/2024/08/23/the-extreme-cost-of-training-ai-models/ (accessed on 20 June 2025).
  96. [96] for Human-Centered Artificial Intelligence (HAI), S.I. AI Index Report. 2024. Available online: https://hai.stanford.edu/ research/ai-index-2024 (accessed on June 2025).
  97. [97] OpenAI. API Pricing, 2025. Available online: https://openai.com/api/pricing/ (accessed on 20 June 2025).
  98. [98] Maslej, N.; Fattorini, L.; Brynjolfsson, E.; Etchemendy, J.; Ligett, K.; Lyons, T.; Manyika, J.; Ngo, H.; Niebles, J.C.; Parli, V.; et al. Artificial intelligence index report 2023. arXiv 2023, arXiv:2310.03715.
  99. [99] Pope, R.; Douglas, S.; Chowdhery, A.; Devlin, J.; Bradbury, J.; Heek, J.; Xiao, K.; Agrawal, S.; Dean, J. Efficiently scaling transformer inference. Proc. Mach. Learn. Syst. 2023, 5, 606–624.
  100. [100] IoT Analytics. The Leading Generative AI Companies. 2025. Available online: https://iot-analytics.com/leading-generative-ai- companies/ (accessed on 19 June 2025).
  101. [101] Zhang, M.; Yuan, B.; Li, H.; Xu, K. LLM-Cloud Complete: Leveraging cloud computing for efficient large language model-based code completion. J. Artif. Intell. Gen. Sci. (JAIGS) 2024, 5, 295–326.
  102. [102] Vipra, J.; Korinek, A. Market concentration implications of foundation models. arXiv 2023, arXiv:2311.01550.
  103. [103] Bommasani, R.; Klyman, K.; Longpre, S.; Kapoor, S.; Maslej, N.; Xiong, B.; Zhang, D.; Liang, P. The 2023 Foundation Model Transparency Index. Trans. Mach. Learn. Res. 2025. https://doi.org/10.48550/arXiv.2310.12941.
  104. [104] Ludwig, J.; Mullainathan, S.; Rambachan, A. Large Language Models: An Applied Econometric Framework; Technical Report; National Bureau of Economic Research: Cambridge, MA, USA, 2025.
  105. [105] Castro, D. AI Can Improve U.S. Small Business Productivity. 2025. Information Technology & Innovation Foundation. Available online: https://itif.org/publications/2025/04/08/ai-can-improve-us-small-business-productivity/ (accessed on 1 Auguest 2025).
  106. [106] Newstardom Insights. The AI Accessibility Gap: Can Small Businesses Keep Up? 2024. Available online: https://newstardom. com/insights/the-ai-accessibility-gap-can-small-businesses-keep-up (accessed on 19 June 2025).
  107. [107] Jain, A.; Kakade, K.S.; Vispute, S.A. The Role of Artificial Intelligence (AI) in the Transformation of Small-and Medium-Sized Businesses: Challenges and Opportunities. In Artificial Intelligence-Enabled Businesses: How to Develop Strategies for Innovation; John Wiley & Sons: Hoboken, NJ, USA, 2025; pp. 209–226.
  108. [108] Korinek, A.; Vipra, J. Concentrating intelligence: Scaling and market structure in artificial intelligence. Econ. Policy 2025, 40, 225–256.
  109. [109] Xie, Y.; Avila, S. The social impact of generative LLM-based AI. Chin. J. Sociol. 2025, 11, 31–57.
  110. [110] Acemoglu, D.; Restrepo, P. Tasks, automation, and the rise in US wage inequality. Econometrica 2022, 90, 1973–2016.
  111. [111] Touvron, H.; Martin, L.; Stone, K.; Albert, P.; Almahairi, A.; Babaei, Y.; Bashlykov, N.; Batra, S.; Bhargava, P.; Bhosale, S.; et al. Llama 2: Open foundation and fine-tuned chat models. arXiv 2023, arXiv:2307.09288.
  112. [112] Grattafiori, A.; Dubey, A.; Jauhri, A.; Pandey, A.; Kadian, A.; Al-Dahle, A.; Letman, A.; Mathur, A.; Schelten, A.; Vaughan, A.; et al. The llama 3 herd of models. arXiv 2024, arXiv:2407.21783.
  113. [113] Phang, J.; Bradley, H.; Gao, L.; Castricato, L.; Biderman, S. EleutherAI: Going Beyond “Open Science” to “Science in the Open”. arXiv 2022, arXiv:2210.06413.
  114. [114] Sinha, S.; Lee, Y.M. Challenges with developing and deploying AI models and applications in industrial systems. Discov. Artif. Intell. 2024, 4, 55.
  115. [115] Fossen, F.; Sorgner, A. Mapping the future of occupations: Transformative and destructive effects of new digital technologies on jobs. Foresight 2019. 13, 10–18.
  116. [116] Wang, Y. The large language model (llm) paradox: Job creation and loss in the age of advanced ai. Authorea Prepr. 2023. https://doi.org/10.36227/techrxiv.24085824.v1.
  117. [117] Durach, C.F.; Gutierrez, L. “Hello, this is your AI co-pilot”–operational implications of artificial intelligence chatbots. Int. J. Phys. Distrib. Logist. Manag. 2024, 54, 229–246.
  118. [118] Dillon, E.W.; Jaffe, S.; Immorlica, N.; Stanton, C.T. Shifting Work Patterns with Generative AI; Technical report; National Bureau of Economic Research: Cambridge, MA, USA, 2025.
  119. [119] Wang, J.Y.; Sukiennik, N.; Li, T.; Su, W.; Hao, Q.; Xu, J.; Huang, Z.; Xu, F.; Li, Y. A Survey on Human-Centric LLMs. arXiv 2024, arXiv:2411.14491.
  120. [120] Niu, Q.; Liu, J.; Bi, Z.; Feng, P.; Peng, B.; Chen, K.; Li, M.; Yan, L.K.; Zhang, Y.; Yin, C.H.; et al. Large language models and cognitive science: A comprehensive review of similarities, differences, and challenges. arXiv 2024, arXiv::2409.02387.
  121. [121] Johnson, S.; Acemoglu, D. Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity|Winners of the 2024 Nobel Prize for Economics; Hachette: London, UK, 2023.
  122. [122] Wilmers, N. Generative AI and the Future of Inequality. MIT Exploration of Generative AI, 27 March 2024. https://doi.org/10.214 28/e4baedd9.777b7123.
  123. [123] Rockall, E.; Mendes Tavares, M.; Pizzinelli, C. AI Adoption and Inequality; International Monetary Fund: Washington, DC, USA, 2025.
  124. [124] Tosun, M.S. Ageing Robots to the Rescue. Technical Report, Oxford Institute of Population Ageing, 2023. Oxford Institute of Population Ageing Blog. Available online: https://www.ageing.ox.ac.uk/blog/ageing-robots-to-the-rescue (accessed on 20 June 2025).
  125. [125] Tosun, M.S. Endogenous fiscal policy and capital market transmissions in the presence of demographic shocks. J. Econ. Dyn. Control 2008, 32, 2031–2060.
  126. [126] Tosun, M.S. Global Aging and Fiscal Policy with International Labor Mobility: A Political Economy Perspective; Technical Report, IZA Discussion Papers; International Monetary Fund: Washington, DC, USA, 2009.
  127. [127] Weeks, W.B.; Spelhaug, J.; Weinstein, J.N.; Ferres, J.M.L. Bridging the rural-urban divide: An implementation plan for leveraging technology and artificial intelligence to improve health and economic outcomes in rural America. J. Rural. Health 2024, 40. https://doi.org/10.1111/jrh.12836.
  128. [128] Woods, D.; Podhorzer, M. AI’s Impact on Income Inequality in the U.S.; Technical Report; Brookings Institution: Washington, DC, USA, 2023.
  129. [129] Acemoglu, D.; Restrepo, P. Secular stagnation? The effect of aging on economic growth in the age of automation. Am. Econ. Rev. 2017, 107, 174–179.
  130. [130] Alayrac, J.B.; Donahue, J.; Luc, P.; Miech, A.; Barr, I.; Hasson, Y.; Lenc, K.; Mensch, A.; Millican, K.; Reynolds, M.; et al. Flamingo: A visual language model for few-shot learning. Adv. Neural Inf. Process. Syst. 2022, 35, 23716–23736.
  131. [131] Shahriar, S.; Lund, B.D.; Mannuru, N.R.; Arshad, M.A.; Hayawi, K.; Bevara, R.V.K.; Mannuru, A.; Batool, L. Putting gpt-4o to the sword: A comprehensive evaluation of language, vision, speech, and multimodal proficiency. Appl. Sci. 2024, 14, 7782.
  132. [132] Islam, R.; Moushi, O.M. Gpt-4o: The cutting-edge advancement in multimodal llm. Authorea Prepr. 2024. https://doi.org/10.100 7/978-3-031-92611-2.
  133. [133] Song, S.; Li, X.; Li, S.; Zhao, S.; Yu, J.; Ma, J.; Mao, X.; Zhang, W.; Wang, M. How to Bridge the Gap between Modalities: Survey on Multimodal Large Language Model. IEEE Trans. Knowl. Data Eng. 2025, 7, 5311–5329.
  134. [134] Wu, J.; Yang, S.; Zhan, R.; Yuan, Y.; Chao, L.S.; Wong, D.F. A survey on LLM-generated text detection: Necessity, methods, and future directions. Comput. Linguist. 2025, 51, 275–338.
  135. [135] Muñoz-Ortiz, A.; Gómez-Rodríguez, C.; Vilares, D. Contrasting linguistic patterns in human and llm-generated news text. Artif. Intell. Rev. 2024, 57, 265.
  136. [136] Dathathri, S.; See, A.; Ghaisas, S.; Huang, P.S.; McAdam, R.; Welbl, J.; Bachani, V.; Kaskasoli, A.; Stanforth, R.; Matejovicova, T.; et al. Scalable watermarking for identifying large language model outputs. Nature 2024, 634, 818–823.
  137. [137] Sun, Y.; He, J.; Cui, L.; Lei, S.; Lu, C.T. Exploring the deceptive power of llm-generated fake news: A study of real-world detection challenges. arXiv 2024, arXiv:2403.18249.
  138. [138] Goswami, A.; Kaur, G.; Tayal, S.; Verma, A.; Verma, M. Analyzing the efficacy of Deep Learning and Transformer models in classifying Human and LLM-Generated Text. In Proceedings of the 2024 8th International Conference on Computing, Communication, Control and Automation (ICCUBEA), Pune, India, 23–24 August 2024; IEEE: New York, NY, USA, 2024; pp. 1–5.
  139. [139] Chan, K.H.; Ke, W.; Im, S.K. A general method for generating discrete orthogonal matrices. IEEE Access 2021, 9, 120380–120391.
  140. [140] Xing, Z.; Lam, C.T.; Yuan, X.; Im, S.K.; Machado, P. Mmqw: Multi-modal quantum watermarking scheme. IEEE Trans. Inf. Forensics Secur. 2024, 19, 5181–5195.
  141. [141] Pu, J.; Sarwar, Z.; Abdullah, S.M.; Rehman, A.; Kim, Y.; Bhattacharya, P.; Javed, M.; Viswanath, B. Deepfake text detection: Limitations and opportunities. In Proceedings of the 2023 IEEE symposium on security and privacy (SP), San Francisco, CA, USA, 21–25 May 2023; IEEE: New York, NY, USA, 2023; pp. 1613–1630.
  142. [142] Wu, J.; Zhan, R.; Wong, D.; Yang, S.; Yang, X.; Yuan, Y.; Chao, L. Detectrl: Benchmarking llm-generated text detection in real-world scenarios. Adv. Neural Inf. Process. Syst. 2024, 37, 100369–100401.
  143. [143] Topsakal, O.; Akinci, T.C. Creating large language model applications utilizing langchain: A primer on developing llm apps fast. In Proceedings of the International Conference on Applied Engineering and Natural Sciences, Konya, Turkey, 10–12 July 2023; Volume 1, pp. 1050–1056.
  144. [144] Cao, S.; Zhang, J.; Shi, J.; Lv, X.; Yao, Z.; Tian, Q.; Li, J.; Hou, L. Probabilistic tree-of-thought reasoning for answering knowledge- intensive complex questions. arXiv 2023, arXiv:2311.13982.
  145. [145] Xu, H.; Zhu, Z.; Pan, L.; Wang, Z.; Zhu, S.; Ma, D.; Cao, R.; Chen, L.; Yu, K. Reducing tool hallucination via reliability alignment. arXiv 2024, arXiv:2412.04141.
  146. [146] Liu, B.; Li, X.; Zhang, J.; Wang, J.; He, T.; Hong, S.; Liu, H.; Zhang, S.; Song, K.; Zhu, K.; et al. Advances and challenges in foundation agents: From brain-inspired intelligence to evolutionary, collaborative, and safe systems. arXiv 2025, arXiv:2504.01990.
  147. [147] Bai, G.; Chai, Z.; Ling, C.; Wang, S.; Lu, J.; Zhang, N.; Shi, T.; Yu, Z.; Zhu, M.; Zhang, Y.; et al. Beyond efficiency: A systematic survey of resource-efficient large language models. arXiv 2024, arXiv:2401.00625.
  148. [148] Weidinger, L.; Mellor, J.; Rauh, M.; Griffin, C.; Uesato, J.; Huang, P.S.; Cheng, M.; Glaese, M.; Balle, B.; Kasirzadeh, A.; et al. Ethical and social risks of harm from language models. arXiv 2021, arXiv:2112.04359.
  149. [149] Epstein, Z.; Hertzmann, A.; of Human Creativity, I.; Akten, M.; Farid, H.; Fjeld, J.; Frank, M.R.; Groh, M.; Herman, L.; Leach, N.; et al. Art and the science of generative AI. Science 2023, 380, 1110–1111.
  150. [150] Desai, D.R.; Riedl, M. Between copyright and computer science: The law and ethics of generative aI. Nw. J. Tech. Intell. Prop. 2024, 22, 55.
  151. [151] Opderbeck, D.W. Copyright in AI training data: A human-centered approach. Okla. L. Rev. 2023, 76, 951.
  152. [152] Raza, S.; Ghuge, S.; Ding, C.; Dolatabadi, E.; Pandya, D. FAIR enough: Develop and assess a FAIR-compliant dataset for large language model training? Data Intell. 2024, 6, 559–585.
  153. [153] Pahune, S.; Akhtar, Z.; Mandapati, V.; Siddique, K. The Importance of AI Data Governance in Large Language Models. Big Data Cogn. Comput. 2025, 9, 147.
  154. [154] Liu, Z. Cultural bias in large language models: A comprehensive analysis and mitigation strategies. J. Transcult. Commun. 2025, 3, 224–244.
  155. [155] Laakso, A. Ethical Challenges of Large Language Models-a Systematic Literature Review. Master’s Thesis, University of Helsinki, Helsinki, Finland, 2023.
  156. [156] Caruana, M.M.; Borg, R.M. Regulating Artificial Intelligence in the European Union. In The EU Internal Market in the Next Decade–Quo Vadis?; Brill: Leiden, The Netherlands, 2025; p. 108.
  157. [157] Fairoze, J.; Garg, S.; Jha, S.; Mahloujifar, S.; Mahmoody, M.; Wang, M. Publicly-detectable watermarking for language models. arXiv 2023, arXiv:2310.18491.
  158. [158] Liu, A.; Pan, L.; Hu, X.; Li, S.; Wen, L.; King, I.; Yu, P.S. An unforgeable publicly verifiable watermark for large language models. arXiv 2023, arXiv:2307.16230.
  159. [159] Li, L.; Bai, Y.; Cheng, M. Where am i from? identifying origin of llm-generated content. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, Miami, FL, USA, 12–16 November 2024; pp. 12218–12229.
  160. [160] Antoun, W.; Sagot, B.; Seddah, D. From text to source: Results in detecting large language model-generated content. arXiv 2023, arXiv:2309.13322.
  161. [161] Russinovich, M.; Salem, A. Hey, That’s My Model! Introducing Chain & Hash, An LLM Fingerprinting Technique. arXiv 2024, arXiv:2407.10887.
  162. [162] Zeng, B.; Wang, L.; Hu, Y.; Xu, Y.; Zhou, C.; Wang, X.; Yu, Y.; Lin, Z. Huref: Human-readable fingerprint for large language models. Adv. Neural Inf. Process. Syst. 2024, 37, 126332–126362.
  163. [163] Bao, G.; Zhao, Y.; Teng, Z.; Yang, L.; Zhang, Y. Fast-detectgpt: Efficient zero-shot detection of machine-generated text via conditional probability curvature. arXiv 2023, arXiv:2310.05130.
  164. [164] Bao, G.; Rong, L.; Zhao, Y.; Zhou, Q.; Zhang, Y. Decoupling Content and Expression: Two-Dimensional Detection of AI-Generated Text. arXiv 2025, arXiv:2503.00258.
  165. [165] Koike, R.; Kaneko, M.; Okazaki, N. Outfox: Llm-generated essay detection through in-context learning with adversarially generated examples. In Proceedings of the AAAI Conference on Artificial Intelligence, Vancouver, BC, Canada, 20–27 February 2024; Volume 38, pp. 21258–21266.
  166. [166] Li, Y.; Li, Q.; Cui, L.; Bi, W.; Wang, L.; Yang, L.; Shi, S.; Zhang, Y. Deepfake text detection in the wild. arXiv 2023, arXiv:2305.13242.
  167. [167] Min, B.; Ross, H.; Sulem, E.; Veyseh, A.P.B.; Nguyen, T.H.; Sainz, O.; Agirre, E.; Heintz, I.; Roth, D. Recent advances in natural language processing via large pre-trained language models: A survey. ACM Comput. Surv. 2023, 56, 1–40.
  168. [168] Shumailov, I.; Shumaylov, Z.; Zhao, Y.; Gal, Y.; Papernot, N.; Anderson, R. The curse of recursion: Training on generated data makes models forget. arXiv 2023, arXiv:2305.17493.
  169. [169] Yang, S.; Chiang, W.L.; Zheng, L.; Gonzalez, J.E.; Stoica, I. Rethinking benchmark and contamination for language models with rephrased samples. arXiv 2023, arXiv:2311.04850.
  170. [170] Li, D.; Sun, R.; Huang, Y.; Zhong, M.; Jiang, B.; Han, J.; Zhang, X.; Wang, W.; Liu, H. Preference Leakage: A Contamination Problem in LLM-as-a-judge. arXiv 2025, arXiv:2502.01534.
  171. [171] Wang, X.; Antoniades, A.; Elazar, Y.; Amayuelas, A.; Albalak, A.; Zhang, K.; Wang, W.Y. Generalization vs Memorization: Tracing Language Models’ Capabilities Back to Pretraining Data. arXiv 2024, arXiv:2407.14985.
  172. [172] Agrawal, A.; Dang, A.; Nezhad, S.B.; Pokharel, R.; Scheinberg, R. Evaluating Multilingual Long-Context Models for Retrieval and Reasoning. arXiv 2024, arXiv:2409.18006.
  173. [173] Ghosh, A.; Datta, D.; Saha, S.; Agarwal, C. The Multilingual Mind: A Survey of Multilingual Reasoning in Language Models. arXiv 2025, arXiv:2502.09457.
  174. [174] Xu, P.; Ping, W.; Wu, X.; McAfee, L.; Zhu, C.; Liu, Z.; Subramanian, S.; Bakhturina, E.; Shoeybi, M.; Catanzaro, B. Retrieval meets long context large language models. In Proceedings of the The Twelfth International Conference on Learning Representations, Kigali, Rwanda, 1–5 May 2023.
  175. [175] Jin, B.; Yoon, J.; Han, J.; Arik, S.O. Long-context llms meet rag: Overcoming challenges for long inputs in rag. arXiv 2024, arXiv:2410.05983.
  176. [176] Villalobos, P.; Ho, A.; Sevilla, J.; Besiroglu, T.; Heim, L.; Hobbhahn, M. Position: Will we run out of data? Limits of LLM scaling based on human-generated data. In Proceedings of the Forty-First International Conference on Machine Learning, Vienna, Austria, 21–27 July 2024.
  177. [177] Nelson, E.; Kollias, G.; Das, P.; Chaudhury, S.; Dan, S. Needle in the haystack for memory based large language models. arXiv 2024, arXiv:2407.01437.
  178. [178] Mohammadabadi, S.M.S. From generative ai to innovative ai: An evolutionary roadmap. arXiv 2025, arXiv:2503.11419.
  179. [179] Dai, H.; Pechi, D.; Yang, X.; Banga, G.; Mantri, R. DENIAHL: In-Context Features Influence LLM Needle-In-A-Haystack Abilities. arXiv 2024, arXiv:2411.19360.
  180. [180] Hsieh, C.P.; Sun, S.; Kriman, S.; Acharya, S.; Rekesh, D.; Jia, F.; Zhang, Y.; Ginsburg, B. RULER: What’s the Real Context Size of Your Long-Context Language Models? arXiv 2024, arXiv:2404.06654.
  181. [181] Yue, X.; Song, Y.; Asai, A.; Kim, S.; de Dieu Nyandwi, J.; Khanuja, S.; Kantharuban, A.; Sutawika, L.; Ramamoorthy, S.; Neubig, G. Pangea: A fully open multilingual multimodal llm for 39 languages. In Proceedings of the The Thirteenth International Conference on Learning Representations, Vienna, Austria, 1–7 May 2024.