Transformers process ID and OOD data differently, enabling effective OOD detection with the BLOOD method.
Abstract
Abstract:
Effective OOD detection is crucial for ML models.
BLOOD leverages transformation smoothness in Transformers for OOD detection.
Outperforms methods with comparable resource requirements.
Introduction:
ML models face challenges with OOD data.
DNNs are popular for OOD detection research.
Different prerequisites categorize OOD detection methods.
Data Extraction:
"Effective out-of-distribution (OOD) detection is crucial for reliable machine learning models."
"We present a novel method for detecting OOD data in Transformers based on transformation smoothness between intermediate layers of a network (BLOOD)."
Quotations:
"The task of discerning between ID and OOD data is commonly referred to as OOD detection."
"Our analysis also suggests that when learning simpler tasks, OOD data transformations maintain their original sharpness."
Out-of-Distribution Detection by Leveraging Between-Layer Transformation Smoothness
Stats
"Effective out-of-distribution (OOD) detection is crucial for reliable machine learning models."
"We present a novel method for detecting OOD data in Transformers based on transformation smoothness between intermediate layers of a network (BLOOD)."
Quotes
"The task of discerning between ID and OOD data is commonly referred to as OOD detection."
"Our analysis also suggests that when learning simpler tasks, OOD data transformations maintain their original sharpness."