Researchers have introduced Activation-Deactivation (AD), a new framework for post-hoc explainable AI that aims to improve the robustness of explanations. Unlike traditional methods that rely on perturbing input features, AD works by deactivating parts of the model itself to simulate the effect of perturbations. An implementation called ConvAD for Convolutional Neural Networks (CNNs) has been developed, which can be integrated into existing trained CNNs without further training. Evaluations across various architectures and datasets demonstrate that ConvAD generates more robust and transferable explanations compared to state-of-the-art methods. AI
IMPACT This new framework could lead to more reliable and trustworthy AI systems by improving the quality of explanations for model behavior.
RANK_REASON The cluster contains an academic paper detailing a new framework for explainable AI. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
