Latest Articles for Out-of-distribution

WeiPer improves out-of-distribution detection in machine learning models using weight adjustments.

Examining how in-distribution labels impact out-of-distribution detection in machine learning.

This article discusses extending context windows in language models using positional vectors.

A new approach to enhance 3D detection accuracy in changing environments.

Exploring safety, reliability, and ethical issues in language models.

A new method helps self-driving cars handle sudden changes safely.

Introducing Ludor, a framework that enhances offline reinforcement learning through knowledge transfer.

A new benchmark suite helps assess reasoning shortcuts in artificial intelligence.

Insights on the challenges of machine learning in predicting material properties.

A new method improves design optimization using existing data.

A new method improves prediction clarity in softmax classifiers for critical fields.

New method enhances visual prediction accuracy through object representation.

Explore the impact of out-of-distribution data on machine learning performance.

LAPT streamlines OOD detection, enhancing AI's reliability in uncertain scenarios.

Introducing ESCAPE, a framework enhancing 3D human pose accuracy and speed.

Research aims to improve AI reliability and energy efficiency in diverse applications.

Test-time augmentation enhances image analysis for gastrointestinal diseases.

Introducing a method to assess AI models on unseen data more effectively.

Exploring how noisy data affects model performance on unseen data.

New methods enhance machine learning models to better detect unusual samples in imbalanced datasets.

Exploring the issues of epistemic uncertainty in Bayesian Deep Learning methods.

Combining existing methods improves OOD detection for safer machine learning applications.

Enhancing accuracy in medical imaging through out-of-distribution detection.

A look at the py-ciu package for explaining AI choices clearly.

New methods enhance predictions of material properties using machine learning techniques.

InfoIGL enhances graph neural networks' performance in varied data environments.

Exploring flexible priors to improve predictions in Bayesian Last Layer models.

SONA creates challenging outliers for better model training in machine learning.

INK provides a reliable method for identifying out-of-distribution samples in machine learning.

A new method integrates human input to enhance OOD learning for machine learning models.

SOOD-ImageNet addresses challenges in computer vision related to changing image meanings.

A new framework enhances object detection by identifying out-of-distribution instances using prototypes.

A new method helps models adapt to unexpected real-world data.

A new approach improves AI's ability to handle unusual data.

A new method simplifies the removal of unwanted content in visual datasets.

Robots learn to handle tricky situations with Object-Centric Recovery.

A graph-based approach to enhance machine learning in dynamic environments.

Unlearning helps AI models forget specific information without losing critical skills.

A new approach to improve OOD detection in machine learning models.

A new method enhances detection of unfamiliar data in deep learning models.