Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 26 for “"Llama-2"”.
-
QLoRaX: Heuristic-Guided Fine-Tuning of LLaMA-2 for Domain Adaptation in Entrepreneurship
… using limited data. We fine-tuned the LLaMA-2 (7B) model on a curated entrepreneurial dataset containing 3,545 human-written question-answer pairs, of which 3,095 were used for training and 450 were reserved for evaluation. A complete fine-tuning and evaluation pipeline was developed, …
-
Velké jazykové modely pro řešení matematických úloh
… open-source modely s 7 miliardami parametrů — Llama-2-7b-hf, Llama-2-7b-chat-hf a Mistral-7B-Instruct-v0.1 — byly doladěny na datové sadě GSM8K pomocí Low-Rank Adaptation (LoRA), se systematickým prohledáváním rychlosti učení, LoRA ranku, alfy a efektivní velikosti dávky. Výkon byl hodnocen …
-
Natūralios kalbos transliavimo į predikatų logikos kalbą galimybės, naudojant didžiuosius kalbos modelius /
… commercial GPT models and the specialized LogicLLaMA solution. The practical contribution consists of fine-tuning two base models from different generations, LLaMA-2-7B and LLaMA-3-8B, using the LoRA method on the MALLS dataset. The results were evaluated using four metrics (SynV, BSC, BLEU, LE) …
-
ADVANCING HEALTH LITERACY THROUGH GENERATIVE AI: THE UTILIZATION OF OPEN-SOURCE LARGE LANGUAGE MODELS (LLMS) FOR TEXT SIMPLIFICATION AND READABILITY
… responses generated by LLMs such as ChatGPT and Llama-2 are evaluated for simplification tasks using the Flesch Kincaid Grade scale. Instruction tuning is performed to evaluate the generation of simplified responses by LLMs according to reading grade level. Fine-tuning of Llama-2-7-B on a dataset …
-
A Novel Approach to Creativity Assessment: Forced Pairwise Ranking With Large Language Models
… Five empirical studies were conducted using Llama-2-7B and Llama-2-13B models adapted via LoRA fine-tuning and benchmarked against human scored responses from the Scientific Creative Thinking Test. A regression baseline achieved Pearson �� = .74 on the test set, matching the human inter-rater …
-
FactPICO : factuality evaluation for plain language summarization of medical evidence
… generated from three LLMs (i.e., GPT-4, Llama-2, and Alpaca), with fine-grained evaluation and natural language rationales from experts. We assess the factuality of critical elements of RCTs in those summaries: Populations, Interventions, Comparators, Outcomes (PICO), as well as the …
-
Teksto lietuvių kalba supaprastinimas į lengvai suprantamą kalbą pasitelkiant didelius kalbos modelius /
… to investigate how the large language model Lt-Llama-2-7b-hf, trained on Lithuanian language data, can be adapted to the task of simplifying Lithuanian texts into Easy-to-Read language. To achieve this goal, a literature analysis was carried out, specialized datasets were developed to fine-tune …
-
Embodied multimodal referring expressions generation
… more data for model training; 5) Finetuning LLaMA 2-chat-13B for generating contextually-correct and situationally-fluent multimodal referring expressions; 6) Integrating the fine-tuned model into the IVA to evaluate the success of the generative model-enabled IVA in communication with …
-
Leveraging Large Language Models for Cybersecurity Risk Assessment of Autonomous Forestry Machines
… 12 demos, and a survey. We used local Llama 2 7B and developed an RAG application by supplying the model with data relevant to cybersecurity and the AGRARSENSE project. We created a generated risk assessment document using the tool. The tool is the main artifact of this study. Also, we …
-
Towards Green AI: Evaluating Energy Efficiency for Quantized LLM Inference
… and runtime. We study quantized inference on Llama 2 7B, Phi 3.5 Mini Instruct, OLMo 1B, Qwen 2.5 7B, and Qwen 2.5 14B across GLUE MNLI, MMLU, HumanEval, and GSM8K datasets, evaluating 10 post-training quantization (PTQ) strategies on NVIDIA H100, A6000, A100, and L4 GPUs. We identify a novel …
-
Enhancing Empathy in Conversational AI with Interpretable Lexicons and Reinforcement Learning from AI Feedback (RLAIF)
… Distilled RLAIF and Direct RLAIF, applied to LLaMA-2-7B-Chat against an established reinforcement learning baseline for emotion-aware empathetic response generation. By combining emotion-lexicon overlap analysis with repetition frequency metrics alongside standard natural language generation …
-
Federated Learning at the Edge for Smart City Sensing Systems
… large language model framework that adapts Llama-2 for long-term time series forecasting. Through cluster-aware federation, channel independence, patching and task alignment techniques, FedTime outperforms strong forecasting baselines across several benchmarks including a production-scale …
-
Statistical Analysis of Outlier Patterns in Low-bit Quantization Strategies in Large Language Models
… We conduct experiments on six model families—Llama-3.1-8B, Llama-3-8B, Llama-2- 7B,Llama-3.2-3B, Mistral-7B-v0.3 and Qwen2.5-7B—across five configurations forming a controlled ablation study: FP16 baseline, weight-only rotation (W4A16KV16), W4A4KV4 without prefix, PrefixQuant-O2 ( η1 = 64), …
-
Toward Reliable Problem Solving with Pre-trained Large Language Models
… in scaled-up versions like GPT-4 and Llama 2, heavily depend on prompt engineering. With well-crafted prompts --- plain text demonstrations --- LLMs are capable of producing coherent language sequences, called thoughts, which serve as intermediate reasoning steps toward solving the …
-
Metodologia para detecção de noticias falsas usando rotulo de vies politico
… (SVM) e para o extrator de características LLAMA 2 como aqueles que obtiveram melhor desempenho, além da proposta se mostrar eficiente tanto para o texto, quanto para o título da notícia.
-
Robustness and Safety in AI Systems: Adaptive Quantile Recalibration for Test-Time Adaptation and Mechanistic Interpretability of LLM Jailbreaking
… attribution patching and subnetwork probing) to LLaMA-2-7B-chat and show that circuits as sparse as ~5% of model edges can be found and attributed to this behavior. Ablating the circuit by zeroing activations at its edges during the forward pass on first token generation, reduces jailbreaking …
-
Explainable Prompt Learning for Movie Review Sentiment Analysis
… a state-of-the-art 98.53% accuracy with the Llama 2 model. We observe that all models achieve stronger results when we apply prompt learning, demonstrating the effectiveness of this approach. In addition to the application of prompting and prompt learning, we also explore the field of …
-
Creation of a Custom Language Model for Pediatric Occupational Therapy Documentation
… point-form notes were used to fine-tune Llama 2 and 3 models on the desired task. Different models’ outputs were evaluated and compared before the final model, a fully fine-tuned Llama 3 8B Instruct model, was selected for a pilot study at KidsAbility in which the custom model was …
-
Aplicación de la inteligencia artificial generativa en la robótica colaborativa para mejorar la eficiencia y flexibilidad en entornos industriales
… modelos de IAG: Claude (en la nube), Mistral-7B, LLama-2-7B, Phi-2 y TinyLLaMA-1.1B (locales), analizando métricas de tiempo de respuesta, precisión de interpretación, diferencial térmico y tiempo de carga inicial mediante análisis estadísticos inferenciales (ANOVA, Chi-cuadrado) para evaluar el …
-
A methodological framework for ontology development, enrichment, and application in natural language processing tasks
… task using "quantized versions" of Llama 2, Flan, Mistral, and Vicuna, four popular LLMs.</p>
Page 1 of 2