Machine-Learning Model as a Subject of Examination
How to identify and document the concrete model artifact, model version, learned parameters, preprocessing, data basis and runtime environment for a technical examination.
Read article
Validating a Machine-Learning Model
How test data, metrics, cross-validation, overfitting and data leakage are examined for a traceable technical assessment of model performance.
Read article
Representativeness of Training and Test Data
How target population, sampling, class and feature distributions, data skew and data drift are examined in training and test data.
Read article
Technical Examination of Bias and Fairness
How groups, selection rates, error rates, demographic parity, equalized odds and limitations are examined in a technical fairness analysis.
Read article
Assessing Machine-Learning Model Robustness
How edge cases, perturbations, stress conditions, performance degradation and adversarial inputs are examined reproducibly.
Read article
Examining LLM Systems with System Prompts, RAG and Tools
How to examine an LLM application as a complete system, including the base model, system prompts, retrieval, knowledge sources, tools, agent permissions, versions and runtime context.
Read article