EvoXplain: Uncovering Divergent Explanations in Machine Learning

This research delves into the critical issue of model explainability, highlighting that even when models achieve similar predictive accuracy, their underlying reasoning can differ significantly. This is important for understanding model behavior and building trust in AI systems.

• •EvoXplain investigates scenarios where ML models agree on predictions but disagree on the underlying reasons.
• •The research analyzes how different training runs can lead to varying internal mechanisms within a model.
• •This work contributes to the development of more transparent and trustworthy AI systems.