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 55 for “"supervised fine-tuning"”.

  1. Foundations of Self-Supervised Fine-Tuning for Open-Weight CLIP Models

    … representations through large-scale self-supervised training on web-scale datasets. CLIP has had a significant impact and is used extensively in downstream vision-language applications. However, improving the representation quality of CLIP models typically requires prohibitively expensive …

    tamu Repository record for Foundations of Self-Supervised Fine-Tuning for Open-Weight CLIP Models (opens in a new tab)

  2. Foundations of Self-Supervised Fine-Tuning for Open-Weight CLIP Models

    … representations through large-scale self-supervised training on web-scale datasets. CLIP has had a significant impact and is used extensively in downstream vision-language applications. However, improving the representation quality of CLIP models typically requires prohibitively expensive …

    tamu Repository record for Foundations of Self-Supervised Fine-Tuning for Open-Weight CLIP Models (opens in a new tab)

  3. Zooming into Comics: Region-Aware RL Improves Fine-Grained Comic Understanding in Vision-Language Models

    … this gap, we introduce AI4VA-FG, the first fine-grained and comprehensive benchmark for VLM-based comic understanding. It spans tasks from foundational recognition and detection to high-level character reasoning and narrative construction, supported by dense annotations for characters, …

    chalmers Repository record for Zooming into Comics: Region-Aware RL Improves Fine-Grained Comic Understanding in Vision-Language Models (opens in a new tab)

  4. Aplikace LLM modelu do oblasti technické diagnostiky a údržby strojů

    … frézky. Ukázka byla realizována pomocí metody Supervised Fine-Tuning (SFT) s maximálním důrazem na lokální řešení.

    brno-tech Repository record for Aplikace LLM modelu do oblasti technické diagnostiky a údržby strojů (opens in a new tab)

  5. Identification of Unknown Landscape Types Using CNN Transfer Learning

    … in critical locations. We propose effective semi-supervised novelty detection approaches for the unknown image type identification problem using Convolutional Neural Network (CNN) Transfer Learning. Recently, the CNN Transfer Learning approach has been very successful in various visual recognition …

    boise Repository record for Identification of Unknown Landscape Types Using CNN Transfer Learning (opens in a new tab)

  6. Complexity Scaling Laws for Neural Models using Combinatorial Optimization

    … model was trained with reinforcement learning or supervised fine-tuning on a static dataset. We conclude with an analogy to problem complexity scaling in local search, showing that a much simpler gradient descent of the cost landscape produces similar trends.

    vt Repository record for Complexity Scaling Laws for Neural Models using Combinatorial Optimization (opens in a new tab)

  7. INDOOR VISION-LANGUAGE NAVIGATION FOR AUTONOMOUS DRONES

    … be adapted for 3D trajectory prediction through supervised fine-tuning, with our best configuration reaching a 9.4% success rate on the evaluation set. This preliminary result is well below the level required for practical deployment; our analysis identifies out-of-distribution handling and the …

    nus Repository record for INDOOR VISION-LANGUAGE NAVIGATION FOR AUTONOMOUS DRONES (opens in a new tab) Download file for INDOOR VISION-LANGUAGE NAVIGATION FOR AUTONOMOUS DRONES (opens in a new tab)

  8. EFFORTS TOWARD TRUSTWORTHY MACHINE LEARNING: MITIGATING OVERCONFIDENCE, HALLUCINATION, AND MODALITY BIAS

    … challenges through post-hoc methods and targeted fine-tuning strategies. First, we mitigate overconfidence in discriminative models by introducing a neighborhood agreement metric that leverages foundation model representations, offering more reliable confidence estimates than softmax scores. We …

    nus Repository record for EFFORTS TOWARD TRUSTWORTHY MACHINE LEARNING: MITIGATING OVERCONFIDENCE, HALLUCINATION, AND MODALITY BIAS (opens in a new tab) Download file for EFFORTS TOWARD TRUSTWORTHY MACHINE LEARNING: MITIGATING OVERCONFIDENCE, HALLUCINATION, AND MODALITY BIAS (opens in a new tab)

  9. ASC-PIE: An Evaluation Framework for PII-Aware Named-Entity Recognition

    … encoder-decoder, and decoder-only models under supervised fine-tuning and in-context prompting. It also proposes SPRINT-PP, a privacy-safe continual-learning method that mitigates forgetting without storing raw historical examples. The evaluation covers extraction quality, robustness, output …

    york Repository record for ASC-PIE: An Evaluation Framework for PII-Aware Named-Entity Recognition (opens in a new tab)

  10. Adaptive self-prompting in agentic LLM frameworks for embedded code fault detection

    … Generation (A-RAG) and Agentic Supervised Fine-Tuning (A-SFT), along with a hybrid architecture integrating both. A-RAG employs confidence-triggered adaptive retrieval on a curated vector knowledge base of coding standards, while A-SFT leverages self-evaluation and instruction …

    uoit Repository record for Adaptive self-prompting in agentic LLM frameworks for embedded code fault detection (opens in a new tab)

  11. Answerability in passage-based question generation by Large Language Models

    … and LLM-generated questions. It evaluates supervised fine-tuning, prompt-based judgment, feature-based modeling, verification-augmented training, and preference adaptation. For detection, Feature-Augmented Peer-Verification Training achieves the strongest open-source performance on both …

    uoit Repository record for Answerability in passage-based question generation by Large Language Models (opens in a new tab)

  12. Enhancing Silicon Sampling in Large Language Models: Fine-Tuning Techniques Using Survey Data

    … sampling. This paper investigates whether supervised fine-tuning can improve algorithmic fidelity in such settings. We construct a dataset from the General Social Survey (GSS), converting demographic attributes into natural-language backstories and randomly selecting one attribute as a …

    byu Repository record for Enhancing Silicon Sampling in Large Language Models: Fine-Tuning Techniques Using Survey Data (opens in a new tab) Download PDF for Enhancing Silicon Sampling in Large Language Models: Fine-Tuning Techniques Using Survey Data (opens in a new tab)

  13. Fine-Grained Text-to-Shape Generation via CLIP Latent Space Adaptation

    … dataset, the supported text descriptions lack fine details. In addition, the CLIP text encoder has a limited token length, which restricts its ability to support rich captions. In this paper, we explore supervised fine-tuning of CLIP-based shape generation with richer captions that enhance …

    carleton Repository record for Fine-Grained Text-to-Shape Generation via CLIP Latent Space Adaptation (opens in a new tab)

  14. Distilling Reasoning from Language Models for Ranking Evaluation

    … distillation from a stronger teacher model, and supervised fine-tuning. These distilled examples are constructed from human-labeled TREC-DL query-document pairs, filtered according to the original annotations, and used to fine-tune open-source models. Experiments were conducted with open-source …

    brazil-ufpe Repository record for Distilling Reasoning from Language Models for Ranking Evaluation (opens in a new tab)

  15. Domain-Aware Reasoning with Lightweight Models via Knowledge Distillation

    … and RAG-retrieved context, the student model was fine-tuned via Supervised Fine-Tuning (SFT) on the teacher’s generated reasoning chains, employing Parameter-Efficient Fine-Tuning (PEFT) with Low-Rank Adaptation (LoRA). Comprehensive evaluation, incorporating both AI-assisted analysis and blind …

    chalmers Repository record for Domain-Aware Reasoning with Lightweight Models via Knowledge Distillation (opens in a new tab)

  16. Evidence-Grounded Biomedical Question Answering: Design, Analysis, and Evaluation

    … A Qwen-based reformulator is adapted through supervised fine-tuning and refined using sequence-level PPO. In the reported BioGen comparison, PPO produces the highest mean downstream reward among the compared reformulation conditions. Overall, the findings show that reliable biomedical question …

    concordia Download PDF for Evidence-Grounded Biomedical Question Answering: Design, Analysis, and Evaluation (opens in a new tab)

  17. Exploring sentence embeddings for better argument relation classification

    … SBERT relies on annotated data, we explore self-supervised fine-tuning of BERT with contrastive learning and various data augmentation strategies to create domain-adaptive sentence embeddings. These embeddings outperform SBERT on text classification tasks when pre-trained and fine-tuned …

    soton Download PDF for Exploring sentence embeddings for better argument relation classification (opens in a new tab)

  18. Aligning language models to professional domains using preference training

    … However, by assuming that a model instruction fine-tuned with labelled data will not be able to outperform a human domain expert, a pairwise comparison dataset can be created from the model's output and the human-generated label, thereby simplifying the training process. These approaches were …

    reykjavik Repository record for Aligning language models to professional domains using preference training (opens in a new tab)

  19. Mitigating Popularity Bias in LLM-Augmented Recommendation Systems Using a Hybrid Framework with a Customized Loss Function and Structured Prompt Engineering

    … under a two-stage continued pre-training and supervised fine-tuning protocol. Lastly, we develop a comprehensive evaluation framework measuring accuracy, popularity bias via Log Popularity Difference (LPD), diversity, and catalog coverage.

    eastern-wash Repository record for Mitigating Popularity Bias in LLM-Augmented Recommendation Systems Using a Hybrid Framework with a Customized Loss Function and Structured Prompt Engineering (opens in a new tab)

  20. Post-Training Large Language Models as Software Engineering Agents

    … space for software engineering agents, spanning supervised fine-tuning (SFT), rejection fine-tuning (RFT), RL from AI feed- back (RLAIF), and RL with verifiable rewards (RLVR). Through three complementary case studies—code question answering (SFT + RLAIF), web-based information retrieval (SFT + …

    waterloo Repository record for Post-Training Large Language Models as Software Engineering Agents (opens in a new tab)

Page 1 of 3