Retrieval strategies (similarity, MMR, hybrid) in Agentic AI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to compute cosine similarity between query and documents.

Agentic AI

similarity_scores = cosine_similarity(query_vector, [1])

Drag options to blanks, or click blank then click option'

Aempty_vectors

Bquery_vector

Crandom_vectors

Ddocument_vectors

Attempts:

3 left

💡 Hint

Common Mistakes

Using the query vector instead of document vectors for comparison.

The cosine similarity compares the query vector to each document vector to find how similar they are.

2fill in blank

medium

Complete the code to select top k documents based on similarity scores.

Agentic AI

top_k_indices = similarity_scores.argsort()[::-1][:[1]]

Drag options to blanks, or click blank then click option'

A0

Bn

Ck

Dlen

Attempts:

3 left

💡 Hint

Common Mistakes

Using 0 or len instead of k for slicing.

We select the top k documents by sorting similarity scores in descending order and taking the first k indices.

3fill in blank

hard

Fix the error in the Maximal Marginal Relevance (MMR) score calculation.

Agentic AI

mmr_score = lambda doc, selected: lambda_param * similarity(doc, query) - (1 - lambda_param) * max(similarity(doc, s) for s in [1])

Drag options to blanks, or click blank then click option'

Aselected

Bquery

Call_docs

Ddocuments

Attempts:

3 left

💡 Hint

Common Mistakes

Using the query or all documents instead of selected documents for redundancy check.

In MMR, we subtract the maximum similarity of the candidate document to already selected documents to reduce redundancy.

4fill in blank

hard

Fill both blanks to implement a hybrid retrieval score combining similarity and MMR.

Agentic AI

hybrid_score = lambda doc, selected: alpha * similarity(doc, query) + (1 - alpha) * [1](doc, selected)
selected_doc = max(candidates, key=lambda doc: [2](doc, selected))

Drag options to blanks, or click blank then click option'

Ammr_score

Bsimilarity

Chybrid_score

Dredundancy_score

Attempts:

3 left

💡 Hint

Common Mistakes

Using similarity instead of mmr_score for the first blank.

Using similarity instead of hybrid_score for selection.

The hybrid score combines similarity and MMR. We use mmr_score function and select the document with the highest hybrid_score.

5fill in blank

hard

Fill all three blanks to create a dictionary of document scores filtered by a threshold.

Agentic AI

filtered_scores = {doc: [1] for doc, score in scores.items() if score [2] [3]

Drag options to blanks, or click blank then click option'

Ascore

B>

Cthreshold

Ddoc

Attempts:

3 left

💡 Hint

Common Mistakes

Using doc instead of score as dictionary value.

Using '<' instead of '>' for filtering.

This dictionary comprehension keeps documents with scores greater than a threshold and maps each document to its score.