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 50 for “"Test-Time Adaptation"”.
-
Test-Time Adaptation with Steering Primitives for Realigning Representations
This work investigates test-time adaptation, the problem of adapting a pretrained model to distribution shift at deployment using only unlabeled test data. Existing methods typically adapt by updating the model’s own parameters, which limits both what can be adapted and where adaptation can occur …
-
TEST-TIME ADAPTATION OF LEARNED IMAGE PRIORS FOR MEDICAL IMAGE RECONSTRUCTION
This thesis studies test-time adaptation of learned image priors for medical image reconstruction under distribution shift. It considers two representative prior types: denoiser-based priors in unrolled reconstruction networks and diffusion priors represented by score networks. For denoiser-based …
-
Improving Video Highlight Detection via Unsupervised Learning and Test-Time Adaptation
… often exhibit limited generalization to unseen test videos, as they rely on generic highlight detection models trained on fixed datasets. This leads to suboptimal performance on unseen test videos due to domain shifts and unique video-specific characteristics not captured during training. To …
-
Time-Series Representation Learning through Dynamic Temporal Reordering and Test-Time Adaptation
Time-series applications span domains such as energy, healthcare, finance, and climate modeling, where the ability to learn rich temporal representations and remain robust under distribution shift is critical. However, existing models often rely on fixed temporal structures that overlook non-local …
-
RecTTA: Reconstruction-based Test-Time Adaptation For Robust Trajectory Prediction In Dynamic Environments
… thesis introduces RecTTA (Reconstruction-based Test-Time Adaptation), a groundbreaking framework that fundamentally transforms trajectory prediction from static inference to dynamic adaptation. Unlike existing domain-specific adaptation methods, RecTTA leverages input trajectory reconstruction …
-
Dealing with domain shift in deep learning: from training-time generalization to test-time adaptation
… representations during training, and Test-Time Adaptation (TTA), which adjusts models dynamically during inference using only unlabeled test data. Together, these approaches address the growing need for reliable and adaptive models in both traditional vision architectures and modern …
-
La vraie utilité de l’adaptation des paramètres des couches de normalisation d’un CNN en Fully Test Time Adaptation
… méthodes qui se basent sur l’approche du Fully Test Time Adaptation, la communauté de la vision par ordinateur apporte un intérêt de recherche de plus en plus important pour étudier cette approche. L’élément caractéristique de cette approche est qu’elle abandonne l’adaptation d’un réseau de …
-
Robustness and Safety in AI Systems: Adaptive Quantile Recalibration for Test-Time Adaptation and Mechanistic Interpretability of LLM Jailbreaking
… models. The first contribution addresses test-time adaptation for image classifiers. In real-world settings, models often face data that differ from the training distribution, leading to performance degradation. To address this, we introduce Adaptive Quantile Recalibration (AQR), a …
-
Backpropagation-free adaptation of vision transformers for robust and efficient 3D and 2D perception
… conditions and assume that the training and test data follow the same distribution. In practice, this assumption is rarely satisfied, especially in 3D industrial applications, where sensors, capture conditions, and objects can vary greatly. These domain shifts lead to significant drops in …
-
Enabling Transfer Learning Across Heterogeneous Domains Using Input Alignment
… to create predictive performance models can be time-consuming. Instead, leveraging existing performance models for predicting the performance of a new application or an existing application on a new architecture is a better use of scientists’ time. However, such knowledge transfer in HPC can be …
-
Computer vision methods adapting to new domains with few samples
… the assumption that the data encountered at test time will be drawn from the same distribution as the training set. In real-world scenarios, this assumption rarely holds. Even subtle variations in lighting, camera properties, background textures, or object appearances can lead to significant …
-
GENERALIZABLE DEEP LEARNING: ADDRESSING DISTRIBUTION SHIFTS FOR VISUAL RECOGNITION
… to distribution shifts between training and test data limits its real-world applicability, particularly in healthcare and autonomous driving. This thesis addresses these challenges by enhancing deep models' generalizability. First, we tackle data distribution shifts through Contrastive …
-
Efficient Multimodal Foundation Model Tuning For Hallucination Mitigation
… lifelong learning. By incorporating test-time adaptation, it seeks to enable multimodal foundation models to evolve through interaction with real data. In this process, an evaluator assesses the overall quality of generated outputs, providing informative feedback that allows the model …
-
ADVANCING ADAPTATION AND ALIGNMENT IN LARGE LANGUAGE MODELS
… several complementary methods. First, we study test-time adaptation with multi-source models that improve through user feedback, proposing Co-UCB, a dueling bandit-based model selection approach. Second, we introduce Preference-Guided Reflective Sampling (PRS), an offline reinforcement learning …
-
OUT-OF-DISTRIBUTION LEARNING FOR ROBUST VISION PERCEPTION
… in deep learning, most models assume that test data match the training distribution (ID). However, real-world data are unpredictable and often contain out-of-distribution (OOD) samples, posing significant challenges. This thesis focuses on data-efficient OOD learning to improve model …
-
Representation Learning under Limited Supervision
… few-shot fine-tuning of foundation models, and adaptation during inference. We propose novel methods to enhance the robustness of semi-supervised learning and few-shot tuning, addressing issues like out-of-distribution data and loss of generalization. Additionally, we unify these approaches into …
-
Test-time fine-tuning via diverse rewards: a conditional variational reward model approach on the abstraction and reasoning corpus
This thesis introduces a framework for test-time adaptation of AI systems that leverages diverse reward signals to refine and optimize performance on challenging tasks, with a focus on the Abstraction and Reasoning Corpus (ARC). At its core, the proposed approach employs a Conditional Variational …
-
Heterogeneous machine learning with decentralized data
Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2026-02-19 without embargo terms
-
Efficient and Test-Time Adaptive Visual Object Tracking in the Wild
… on in-distribution (ID) datasets, i.e., on the testing portion of the same datasets used for training. However, they incur high computational costs and hardware constraints, making their deployment in the wild for mobile, autonomous, and IoT applications still challenging. Efficient visual …
-
A Robust Learning Framework for Resource-Constrained Domain Adaptation
… machine learning models assume that training and test data are independently and identically distributed (\iid) according to the same data distribution. Under this assumption, a model trained on labeled data is expected to generalize well to unseen test data. However, real-world applications often …
Page 1 of 3