Retrieval strategies (similarity, MMR, hybrid) in Agentic AI - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is a similarity-based retrieval strategy?

A similarity-based retrieval strategy finds items that are most alike to a query by measuring how close or similar their features are. It often uses metrics like cosine similarity or Euclidean distance to rank results.

intermediate

Explain Maximal Marginal Relevance (MMR) in retrieval.

MMR balances relevance and diversity by selecting items that are both similar to the query and different from each other. This helps avoid redundant results and provides a varied set of answers.

intermediate

What does a hybrid retrieval strategy combine?

A hybrid retrieval strategy combines similarity-based methods with other techniques like MMR or rule-based filters to improve both relevance and diversity of results.

beginner

Why is diversity important in retrieval results?

Diversity ensures that retrieved results cover different aspects or perspectives, reducing repetition and increasing the chance of finding useful information.

beginner

Name two common similarity metrics used in retrieval strategies.

Cosine similarity and Euclidean distance are two common metrics used to measure how close or similar items are in retrieval tasks.

Which retrieval strategy explicitly balances relevance and diversity?

A Pure similarity-based retrieval

B Maximal Marginal Relevance (MMR)

C Random selection

D Rule-based filtering

MMR selects items that are relevant to the query but also different from each other to ensure diversity.

What does a similarity-based retrieval strategy primarily use to rank results?

A Predefined categories

B Random chance

C User feedback only

D Similarity metrics like cosine similarity

Similarity-based retrieval ranks items by how close they are to the query using similarity metrics.

A hybrid retrieval strategy might combine similarity with which other approach?

A Only random selection

B Ignoring diversity

C Maximal Marginal Relevance (MMR)

D No filtering

Hybrid strategies combine similarity with methods like MMR to improve result quality.

Why avoid only similarity-based retrieval without diversity?

A Results may be repetitive and less useful

B It is faster

C It always gives the best results

D It ignores relevance

Without diversity, results can be very similar and redundant, reducing usefulness.

Which metric measures the angle between two vectors in retrieval?

A Cosine similarity

B Euclidean distance

C Manhattan distance

D Jaccard index

Cosine similarity measures the angle between vectors, indicating how similar they are.

Describe how Maximal Marginal Relevance (MMR) improves retrieval results compared to pure similarity-based methods.

Think about why showing only very similar items might not be helpful.

Explain what a hybrid retrieval strategy is and why it might be better than using only one retrieval method.

Consider mixing strengths of different approaches.