Multi-query retrieval for better recall in LangChain - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ component_behavior

intermediate

What is the output of this multi-query retrieval code snippet?

Consider this LangChain code that performs multi-query retrieval to improve recall. What will be the printed output?

LangChain

from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
from langchain.chains import RetrievalQA
from langchain.llms import OpenAI

embeddings = OpenAIEmbeddings()
vectorstore = FAISS.load_local("my_faiss_index", embeddings)

retriever = vectorstore.as_retriever(search_kwargs={"k": 2})

queries = ["What is AI?", "Explain machine learning."]

results = []
for query in queries:
    qa = RetrievalQA.from_chain_type(llm=OpenAI(), chain_type="stuff", retriever=retriever)
    answer = qa.run(query)
    results.append(answer)

print(results)

A A list of two answer strings, each answering one query

B A single string combining answers to both queries

C A list of two empty strings

D Raises a TypeError because 'retriever' is not iterable

💡 Hint

Think about how the loop runs the retrieval QA chain separately for each query and collects answers.

The code runs a retrieval QA chain for each query separately and collects the answers in a list. So the output is a list of two answer strings, one per query.

📝 Syntax

intermediate

Which option correctly initializes a multi-query retriever with LangChain?

Choose the correct code snippet that sets up a retriever to handle multiple queries with LangChain's FAISS vector store.

A retriever = vectorstore.as_retriever(k=3)

B retriever = vectorstore.as_retriever(search_k=3)

C retriever = vectorstore.as_retriever({"k": 3})

D retriever = vectorstore.as_retriever(search_kwargs={"k": 3})

💡 Hint

Check the correct parameter name for passing search options in LangChain retrievers.

The correct way is to pass search parameters inside a dictionary to the 'search_kwargs' argument. So option D is correct.

🔧 Debug

advanced

Why does this multi-query retrieval code raise an error?

Given this code snippet, why does it raise an error?

LangChain

from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings

embeddings = OpenAIEmbeddings()
vectorstore = FAISS.load_local("index_dir", embeddings)

queries = ["Define AI", "What is NLP?"]

retriever = vectorstore.as_retriever(search_kwargs={"k": 2})

answers = [retriever.run(query) for query in queries]
print(answers)

A TypeError because queries is not iterable

B SyntaxError due to missing colon

C 'Retriever' object has no attribute 'run' error

D No error, prints list of answers

💡 Hint

Check if the retriever object supports a 'run' method.

The retriever object does not have a 'run' method. It is used to fetch documents, not to answer queries directly. You need to use a chain like RetrievalQA to get answers.

❓ state_output

advanced

What is the value of 'combined_answers' after running this multi-query retrieval code?

Analyze the code and determine the value of the variable combined_answers after execution.

LangChain

from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
from langchain.chains import RetrievalQA
from langchain.llms import OpenAI

embeddings = OpenAIEmbeddings()
vectorstore = FAISS.load_local("faiss_index", embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 1})

queries = ["Explain AI", "What is deep learning?"]

answers = []
for q in queries:
    qa_chain = RetrievalQA.from_chain_type(llm=OpenAI(), chain_type="stuff", retriever=retriever)
    answers.append(qa_chain.run(q))

combined_answers = " | ".join(answers)

A A list of empty strings

B A single string combining two answers separated by ' | '

C An empty string

D A list of two answer strings

💡 Hint

Look at how the answers list is joined into a single string.

The answers list contains two strings. Joining them with ' | ' creates one combined string with answers separated by ' | '.

🧠 Conceptual

expert

Why use multi-query retrieval in LangChain for better recall?

Which reason best explains why multi-query retrieval improves recall in LangChain applications?

A It allows querying multiple related questions separately to gather more relevant documents

B It speeds up retrieval by running queries in parallel

C It reduces the number of documents retrieved to save memory

D It merges all queries into one to simplify processing

💡 Hint

Think about how splitting a complex question into parts affects document retrieval.

Multi-query retrieval splits a complex information need into multiple queries, each retrieving relevant documents. This increases the chance of finding all useful information, improving recall.