Tag: adversarial machine learning
Browse our exclusive articles!
H-Sets: Discovering Feature Interactions in Image Classifiers
H-Sets framework enhances image classifier interpretability by detecting and attributing set-level feature interactions using Hessian-guided methods.
Memory Tokens Boost Universal Transformer Performance
Explore how memory tokens improve Universal Transformers' adaptive reasoning, enhancing accuracy and efficiency in complex tasks like Sudoku solving.
Why Large Language Models Fail at Random Number Sampling
Discover why large language models struggle with generating random numbers from statistical distributions and the impact on AI applications.
Robust LLM-Based Math Reasoning Evaluation Framework
Discover a novel LLM-based framework that improves math reasoning evaluation, surpassing traditional symbolic methods with enhanced accuracy and flexibilit...
Background Temperature Reveals Hidden Randomness in LLMs
Discover how background temperature explains hidden randomness in large language models, improving reproducibility and evaluation.