1Z0-1122-25 Cheatsheet — AI Fundamentals, Metrics, GenAI Basics & Responsible AI

Last-mile 1Z0-1122-25 review: AI/ML lifecycle, evaluation metrics pickers, leakage/overfitting rules, GenAI grounding intuition, and responsible AI checklists.

On this page

Use this for last‑mile review. Pair it with the Syllabus.

1) The AI/ML lifecycle (what every scenario maps to)

    flowchart LR
	  P["Problem framing"] --> D["Data + labels"]
	  D --> F["Features"]
	  F --> T["Train"]
	  T --> E["Evaluate"]
	  E --> DEP["Deploy"]
	  DEP --> MON["Monitor + iterate"]

Exam cue: if you skip evaluation/monitoring, the option is usually incomplete.

2) Metrics pickers (high-yield)

Task	Good default	When to change
Classification	F1 / AUC	use precision/recall when FP/FN costs differ
Regression	MAE / RMSE	RMSE punishes large errors more

Rule: If the prompt mentions class imbalance, accuracy is rarely the best answer.

3) Data pitfalls (the “why did the model fail?” list)

Leakage: features include future information → unrealistically good offline metrics.
Overfitting: train metrics great, test metrics poor → simplify model/regularize/more data.
Label noise: inconsistent labels → fix labeling process before tuning models.

4) GenAI basics (what’s actually being tested)

Concept	What it means	Practical implication
Tokens	text pieces	cost and latency scale with tokens
Context window	max prompt + docs	long docs require chunking
Hallucination	plausible but wrong	add grounding + citations

5) “Grounding” intuition (RAG in one picture)

    flowchart LR
	  Q["Question"] --> RET["Retrieve relevant docs"]
	  RET --> PROMPT["Prompt with context"]
	  PROMPT --> LLM["LLM"]
	  LLM --> A["Answer + citations"]

Rule: grounded answers come from good retrieval + clean data, not clever prompts.

6) Responsible AI checklist (exam-friendly)

Bias/fairness: evaluate across segments; watch for proxy features.
Privacy: minimize sensitive data; control access; avoid training on secrets.
Security: prompt injection awareness; validate inputs; least privilege.
Transparency: document data sources, limitations, and monitoring signals.

Syllabus

Practice

Browse Exams — Mock Exams & Practice Tests