ManySpace
Türkçe Sign in Sign up

AI & MLRecommendedAIML-RET-012

Evaluating Retrieval-Augmented Generation

Separating retrieval quality from generation quality: context recall, faithfulness, and answer correctness.

Difficulty
3 / 5
Importance
4 / 5
Study time
1–2 h
Kind of work
Mixed

Why it matters

The learner can diagnose a failed RAG answer as a retrieval miss or a generation error using component metrics.

The route to this topic

Route AIML-RET-012
What Is a Model? Learning from Data Tasks: Regression, Classification, and Beyond Features and Targets Datasets and Splits Classification Framing Linear Regression I: Prediction and Residuals Embeddings Tokens and Tokenization Train, Validation, and Test Errors Scores and Logits Loss and Fitting: Squared Error Similarity in Vector Space Offline Evaluation Multiclass Scores and Softmax Autoregression Objective Functions and Empirical Risk Metric Spaces and Distance Query, Key, Value Semantic Embeddings Ranking Metrics Gradient Descent in Machine Learning Likelihood Next-Token Prediction Similarity Scores Similarity Search Chunking Stochastic and Mini-batch Optimization Cross-Entropy Loss Generation Metrics Weighted Aggregation Retrieval The Language-Modeling Objective Self-Attention Pretraining Context and Conditioning LLM Evaluation Retrieval-Augmented Generation Grounding and Citation Evaluating Retrieval-Augmented Generation