Evaluation & AI Reliability · core
Model Evaluation
Held-out loss, baselines, hallucination, leakage.
Mental model
Offline eval (on a held-out test set) tells you the model got better. Online eval (A/B test on real users) tells you the product got better. You need both — neither alone is enough. Scope: this card owns the offline/online split and the classical failure modes — held-out sets, baselines, leakage, distribution shift. Building a grader for free-text output is `llm-evals`; comparing whole systems on cost and latency is `quality-cost-latency-measurement`.
How to study Model Evaluation
Begin by restating the mental model in your own words, then connect it to a concrete system you have built or operated. Name the mechanism, the constraint it addresses, and the trade-off it introduces. Use Stanford CS336 — L12 Evaluation (Spring 2025), Eugene Yan — Evaluation & Hallucination, Your AI product needs evals (Hamel Husain) to check details, but close the source before writing your explanation. Retrieval is the learning step; rereading is only preparation.
Next, compare Model Evaluation with LLM Evals. Ask what changes in correctness, latency, resource use, operability, and failure recovery. Complete Perplexity vs classification accuracy and preserve the command, input, output, and one failed attempt as evidence. Finish by explaining the idea without jargon to someone who has not studied the track.
Proof of understanding
- Explain the mechanism from first principles and identify the state it reads or changes.
- Give one situation where the concept is the right choice and one where it is not.
- Predict a realistic failure mode before running the drill, then compare the prediction with evidence.
- Connect the result to a roadmap or build artifact instead of treating the concept as isolated trivia.
Learn from primary sources
Practice and explain it back
Perplexity vs classification accuracy
LM benchmark: PPX dropped 5% but exact-match QA accuracy flat. What does that imply about the eval?
Expected evidence: LM loss improved token distribution but task metric insensitive or needs few-shot/prompt — check task-specific eval.
Open the interactive drill →Review prompts
- Held-out loss improved but the A/B test was flat. Name two explanations that are not "the metric is wrong".
Build evidence
Synthesize: AI Models & Training
Move from transformer foundations through pre-training, fine-tuning, post-training, compression, and evaluation. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks
Synthesize: Evaluation & AI Reliability
Build an evidence-backed evaluation and observability system for models, tools, and agents. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks