According to best practices for MLOps, what must be version-controlled alongside model weights and training data?
Answer
The methodology used to generate the explanations
Best practices dictate that explanations must be treated as auditable artifacts and version-controlled, just like weights and data, to mitigate the risk of 'explanation drift' when the model updates.
#Videos
What Is Explainable AI? - YouTube
Related Questions
What is the primary goal of Artificial Intelligence explainability (XAI)?What term describes the core challenge presented by the most powerful AI models?What distinguishes Interpretability from Explainability?Which model type is explicitly mentioned as being inherently interpretable due to its clear, mathematically traceable equation?What question does Global Explainability seek to answer?Which scope of explanation is most vital for end-users or auditors needing to contest an immediate outcome, such as a denied loan?Which post-hoc technique, rooted in cooperative game theory, calculates the unique contribution of each feature relative to the average prediction?What visualization tool highlights the specific pixels or words a neural network paid the most "attention" to when making an image classification decision?According to best practices for MLOps, what must be version-controlled alongside model weights and training data?What security-related risk can attackers exploit by using explanation methods to reverse-engineer decision boundaries?Which stakeholder group primarily desires high-level summaries of key drivers and actionable insights into performance drift?