Accuracy describes how often an AI model produces correct results; stability describes how reliably that performance holds up – with slightly altered inputs, new data or over time. A model can be very accurate in testing yet unstable if merely rephrasing a question leads to a different answer. For real-world use, both properties are checked with fixed test cases, cross-validation and ongoing monitoring.
Model robustness describes how reliably an AI model performs when inputs are noisy, incomplete, unusual or deliberately manipulated. A robust model still delivers stable results, such as image recognition in poor lighting or a language model handling typos. Robustness is assessed with stress tests, diverse training data and tests against deliberate deception attempts, and it is particularly important in safety-critical applications.
Model drift is the gradual decline in a model’s predictive quality during operation because reality changes compared with the training data. A distinction is made between changes in the input data (data drift) and changes in the relationship between inputs and outcomes (concept drift). Example: a fraud detection model misses new fraud schemes. Remedies include continuous monitoring of metrics and regular retraining.
Cross-validation is a method for checking how well a model generalises to new, unseen data. In k-fold cross-validation, the data is split into k parts; the model is trained k times on k−1 parts and tested on the remaining part each time, after which the results are averaged. This helps detect overfitting and compare models or settings fairly, especially when little data is available.
In AI, benchmarking means systematically comparing models or systems using defined tasks and criteria. Public benchmarks measure things like general knowledge, reasoning or coding ability; for practical use, however, test cases from your own day-to-day work matter most. Typical criteria are accuracy, completeness, response time and cost. This helps decide whether a smaller, cheaper model is good enough for a task.
