Evaluation of RAG performance basics

This tutorial shows how to use Fabric to evaluate RAG application performance. The evaluation focuses on two main RAG components: the retriever (Azure AI Search) and the response generator (an LLM that uses the user's query, retrieved context, and a prompt to generate a reply). Here are the main steps:

Set up Azure OpenAI and Azure AI Search services
Load data from CMU's QA dataset of Wikipedia articles to build a benchmark
Run a smoke test with one query to confirm the RAG system works end to end
Define deterministic and AI-assisted metrics for evaluation
Check-in 1: Evaluate retriever performance using top-N accuracy
Check-in 2: Evaluate response generator performance using groundedness, relevance, and similarity metrics
Visualize and store evaluation results in OneLake for future reference and ongoing evaluation

Prerequisites

Before you start this tutorial, complete the Building Retrieval Augmented Generation in Fabric step-by-step guide.

You need these services to run the notebook:

Microsoft Fabric
Add a lakehouse to this notebook (it contains the data you added in the previous tutorial).
Azure AI Studio for OpenAI
Azure AI Search (it contains the data you indexed in the previous tutorial).

In the previous tutorial, you uploaded data to your lakehouse and built a document index used by the RAG system. Use the index in this exercise to learn core techniques to evaluate RAG performance and identify potential problems. If you didn't create an index or removed it, follow the quickstart guide to complete the prerequisite.

Set up access to Azure OpenAI and Azure AI Search

Define endpoints and required keys. Import required libraries and functions. Instantiate clients for Azure OpenAI and Azure AI Search. Define a function wrapper with a prompt to query the RAG system.

# Enter your Azure OpenAI service values
aoai_endpoint = "https://<your-resource-name>.openai.azure.com" # TODO: Provide the Azure OpenAI resource endpoint (replace <your-resource-name>)
aoai_key = "" # TODO: Fill in your API key from Azure OpenAI 
aoai_deployment_name_embeddings = "text-embedding-ada-002"
aoai_model_name_query = "gpt-4-32k"  
aoai_model_name_metrics = "gpt-4-32k"
aoai_api_version = "2024-02-01"

# Setup key accesses to Azure AI Search
aisearch_index_name = "" # TODO: Create a new index name: must only contain lowercase, numbers, and dashes
aisearch_api_key = "" # TODO: Fill in your API key from Azure AI Search
aisearch_endpoint = "https://.search.windows.net" # TODO: Provide the url endpoint for your created Azure AI Search

import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning) 

import os, requests, json

from datetime import datetime, timedelta
from azure.core.credentials import AzureKeyCredential
from azure.search.documents import SearchClient

from pyspark.sql import functions as F
from pyspark.sql.functions import to_timestamp, current_timestamp, concat, col, split, explode, udf, monotonically_increasing_id, when, rand, coalesce, lit, input_file_name, regexp_extract, concat_ws, length, ceil
from pyspark.sql.types import StructType, StructField, StringType, IntegerType, TimestampType, ArrayType, FloatType
from pyspark.sql import Row
import pandas as pd
from azure.search.documents.indexes import SearchIndexClient
from azure.search.documents.models import (
    VectorizedQuery,
)
from azure.search.documents.indexes.models import (  
    SearchIndex,  
    SearchField,  
    SearchFieldDataType,  
    SimpleField,  
    SearchableField,   
    SemanticConfiguration,  
    SemanticPrioritizedFields,
    SemanticField,  
    SemanticSearch,
    VectorSearch,  
    HnswAlgorithmConfiguration,
    HnswParameters,  
    VectorSearchProfile,
    VectorSearchAlgorithmKind,
    VectorSearchAlgorithmMetric,
)

import openai 
from openai import AzureOpenAI
import uuid
import matplotlib.pyplot as plt
from synapse.ml.featurize.text import PageSplitter
import ipywidgets as widgets  
from IPython.display import display as w_display

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 6, Finished, Available, Finished)

# Configure access to OpenAI endpoint
openai.api_type = "azure"
openai.api_key = aoai_key
openai.api_base = aoai_endpoint
openai.api_version = aoai_api_version

# Create client for accessing embedding endpoint
embed_client = AzureOpenAI(
    api_version=aoai_api_version,
    azure_endpoint=aoai_endpoint,
    api_key=aoai_key,
)

# Create client for accessing chat endpoint
chat_client = AzureOpenAI(
    azure_endpoint=aoai_endpoint,
    api_key=aoai_key,
    api_version=aoai_api_version,
)

# Configure access to Azure AI Search
search_client = SearchClient(
    aisearch_endpoint,
    aisearch_index_name,
    credential=AzureKeyCredential(aisearch_api_key)
)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 7, Finished, Available, Finished)

The following functions implement the two main RAG components - retriever (get_context_source) and response generator (get_answer). The code is similar to the previous tutorial. The topN parameter lets you set how many relevant resources to retrieve (this tutorial uses 3, but the optimal value can vary by dataset):

# Implement retriever
def get_context_source(question, topN=3):
    """
    Retrieves contextual information and sources related to a given question using embeddings and a vector search.  
    Parameters:  
    question (str): The question for which the context and sources are to be retrieved.  
    topN (int, optional): The number of top results to retrieve. Default is 3.  
      
    Returns:  
    List: A list containing two elements:  
        1. A string with the concatenated retrieved context.  
        2. A list of retrieved source paths.  
    """
    embed_client = openai.AzureOpenAI(
        api_version=aoai_api_version,
        azure_endpoint=aoai_endpoint,
        api_key=aoai_key,
    )

    query_embedding = embed_client.embeddings.create(input=question, model=aoai_deployment_name_embeddings).data[0].embedding

    vector_query = VectorizedQuery(vector=query_embedding, k_nearest_neighbors=topN, fields="Embedding")

    results = search_client.search(   
        vector_queries=[vector_query],
        top=topN,
    )

    retrieved_context = ""
    retrieved_sources = []
    for result in results:
        retrieved_context += result['ExtractedPath'] + "\n" + result['Chunk'] + "\n\n"
        retrieved_sources.append(result['ExtractedPath'])

    return [retrieved_context, retrieved_sources]

# Implement response generator
def get_answer(question, context):
    """  
    Generates a response to a given question using provided context and an Azure OpenAI model.  
    
    Parameters:  
        question (str): The question that needs to be answered.  
        context (str): The contextual information related to the question that will help generate a relevant response.  
    
    Returns:  
        str: The response generated by the Azure OpenAI model based on the provided question and context.  
    """
    messages = [
        {
            "role": "system",
            "content": "You are a chat assistant. Use provided text to ground your response. Give a one-word answer when possible ('yes'/'no' is OK where appropriate, no details). Unnecessary words incur a $500 penalty."
        }
    ]

    messages.append(
        {
            "role": "user", 
            "content": question + "\n" + context,
        },
    )

    chat_client = openai.AzureOpenAI(
        azure_endpoint=aoai_endpoint,
        api_key=aoai_key,
        api_version=aoai_api_version,
    )

    chat_completion = chat_client.chat.completions.create(
        model=aoai_model_name_query,
        messages=messages,
    )

    return chat_completion.choices[0].message.content

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 8, Finished, Available, Finished)

Dataset

Version 1.2 of the Carnegie Mellon University Question-Answer dataset is a corpus of Wikipedia articles with factual questions and answers written manually. It's hosted in Azure Blob Storage under the GFDL. The dataset uses one table with these fields:

ArticleTitle: Name of the Wikipedia article the questions and answers come from
Question: Manually written question about the article
Answer: Manually written answer based on the article
DifficultyFromQuestioner: Difficulty rating the question author assigns
DifficultyFromAnswerer: Difficulty rating the evaluator assigns; can differ from DifficultyFromQuestioner
ExtractedPath: Path to the original article (an article can have multiple question-answer pairs)
text: Cleaned Wikipedia article text

Download the LICENSE-S08 and LICENSE-S09 files from the same location for license details.

History and citation

Use this citation for the dataset:

CMU Question/Answer Dataset, Release 1.2
August 23, 2013
Noah A. Smith, Michael Heilman, and Rebecca Hwa
Question Generation as a Competitive Undergraduate Course Project
In Proceedings of the NSF Workshop on the Question Generation Shared Task and Evaluation Challenge, Arlington, VA, September 2008. 
Available at http://www.cs.cmu.edu/~nasmith/papers/smith+heilman+hwa.nsf08.pdf.
Original dataset acknowledgments:
This research project was supported by NSF IIS-0713265 (to Smith), an NSF Graduate Research Fellowship (to Heilman), NSF IIS-0712810 and IIS-0745914 (to Hwa), and Institute of Education Sciences, U.S. Department of Education R305B040063 (to Carnegie Mellon).
cmu-qa-08-09 (modified version)
June 12, 2024
Amir Jafari, Alexandra Savelieva, Brice Chung, Hossein Khadivi Heris, Journey McDowell
This release uses the GNU Free Documentation License (GFDL) (http://www.gnu.org/licenses/fdl.html).
The GNU license applies to all copies of the dataset.

Create benchmark

Import the benchmark. For this demo, use a subset of questions from the S08/set1 and S08/set2 buckets. To keep one question per article, apply df.dropDuplicates(["ExtractedPath"]). Drop duplicate questions. The curation process adds difficulty labels; this example limits them to medium.

df = spark.sql("SELECT * FROM data_load_tests.cmu_qa")

# Filter the DataFrame to include the specified paths
df = df.filter((col("ExtractedPath").like("S08/data/set1/%")) | (col("ExtractedPath").like("S08/data/set2/%")))

# Keep only medium-difficulty questions.
df = df.filter(col("DifficultyFromQuestioner") == "medium")


# Drop duplicate questions and source paths.
df = df.dropDuplicates(["Question"])
df = df.dropDuplicates(["ExtractedPath"])

num_rows = df.count()
num_columns = len(df.columns)
print(f"Number of rows: {num_rows}, Number of columns: {num_columns}")

# Persist the DataFrame
df.persist()
display(df)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 9, Finished, Available, Finished) Number of rows: 20, Number of columns: 7 SynapseWidget(Synapse.DataFrame, 47aff8cb-72f8-4a36-885c-f4f3bb830a91)

The result is a DataFrame with 20 rows - the demo benchmark. Key fields are Question, Answer (human-curated ground truth answer), and ExtractedPath (the source document). Adjust the filters to include other questions and vary difficulty for a more realistic example. Try it.

Run a simple end-to-end test

Start with an end-to-end smoke test of retrieval-augmented generation (RAG).

question = "How many suborders are turtles divided into?"
retrieved_context, retrieved_sources = get_context_source(question)
answer = get_answer(question, retrieved_context)
print(answer)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 10, Finished, Available, Finished) Three

This smoke test helps you find issues in the RAG implementation, such as incorrect credentials, a missing or empty vector index, or incompatible function interfaces. If the test fails, check for issues. Expected output: Three. If the smoke test passes, go to the next section to evaluate RAG further.

Establish metrics

Define a deterministic metric to evaluate the retriever. It's inspired by search engines. It checks whether the retrieved sources list includes the ground truth source. This metric is a top-N accuracy score because the topN parameter sets the number of retrieved sources.

def get_retrieval_score(target_source, retrieved_sources):
    if target_source in retrieved_sources: 
        return 1
    else: 
        return 0

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 11, Finished, Available, Finished)

According to the benchmark, the answer is contained in the source with ID "S08/data/set1/a9". Testing the function on the example we ran above returns 1, as expected, because it was in the top three relevant text chunks.

print("Retrieved sources:", retrieved_sources)
get_retrieval_score("S08/data/set1/a9", retrieved_sources)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 12, Finished, Available, Finished) Retrieved sources: ['S08/data/set1/a9', 'S08/data/set1/a9', 'S08/data/set1/a5'] 1

This section defines AI-assisted metrics. The prompt template includes a few examples of input (CONTEXT and ANSWER) and suggested output - also known as a few-shot model. It's the same prompt that's used in Azure AI Studio. Learn more in Built-in evaluation metrics. This demo uses the groundedness and relevance metrics - these are usually the most useful and reliable for evaluating GPT models. Other metrics can be useful but provide less intuition - for example, answers don't have to be similar to be correct, so similarity scores can be misleading. The scale for all metrics is 1 to 5. Higher is better. Groundedness takes only two inputs (context and generated answer), while the other two metrics also use ground truth for evaluation.

def get_groundedness_metric(context, answer):
    """Get the groundedness score from the LLM using the context and answer."""

    groundedness_prompt_template = """
    You are presented with a CONTEXT and an ANSWER about that CONTEXT. Decide whether the ANSWER is entailed by the CONTEXT by choosing one of the following ratings:
    1. 5: The ANSWER follows logically from the information contained in the CONTEXT.
    2. 1: The ANSWER is logically false from the information contained in the CONTEXT.
    3. an integer score between 1 and 5 and if such integer score does not exist, use 1: It is not possible to determine whether the ANSWER is true or false without further information. Read the passage of information thoroughly and select the correct answer from the three answer labels. Read the CONTEXT thoroughly to ensure you know what the CONTEXT entails. Note the ANSWER is generated by a computer system, it can contain certain symbols, which should not be a negative factor in the evaluation.
    Independent Examples:
    ## Example Task #1 Input:
    "CONTEXT": "Some are reported as not having been wanted at all.", "QUESTION": "", "ANSWER": "All are reported as being completely and fully wanted."
    ## Example Task #1 Output:
    1
    ## Example Task #2 Input:
    "CONTEXT": "Ten new television shows appeared during the month of September. Five of the shows were sitcoms, three were hourlong dramas, and two were news-magazine shows. By January, only seven of these new shows were still on the air. Five of the shows that remained were sitcoms.", "QUESTION": "", "ANSWER": "At least one of the shows that were cancelled was an hourlong drama."
    ## Example Task #2 Output:
    5
    ## Example Task #3 Input:
    "CONTEXT": "In Quebec, an allophone is a resident, usually an immigrant, whose mother tongue or home language is neither French nor English.", "QUESTION": "", "ANSWER": "In Quebec, an allophone is a resident, usually an immigrant, whose mother tongue or home language is not French."
    5
    ## Example Task #4 Input:
    "CONTEXT": "Some are reported as not having been wanted at all.", "QUESTION": "", "ANSWER": "All are reported as being completely and fully wanted."
    ## Example Task #4 Output:
    1
    ## Actual Task Input:
    "CONTEXT": {context}, "QUESTION": "", "ANSWER": {answer}
    Reminder: The return values for each task should be correctly formatted as an integer between 1 and 5. Do not repeat the context and question.  Don't explain the reasoning. The answer should include only a number: 1, 2, 3, 4, or 5.
    Actual Task Output:
    """

    metric_client = openai.AzureOpenAI(
        api_version=aoai_api_version,
        azure_endpoint=aoai_endpoint,
        api_key=aoai_key,
    )

    messages = [
        {
            "role": "system",
            "content": "You are an AI assistant. You will be given the definition of an evaluation metric for assessing the quality of an answer in a question-answering task. Your job is to compute an accurate evaluation score using the provided evaluation metric."
        }, 
        {
            "role": "user",
            "content": groundedness_prompt_template.format(context=context, answer=answer)
        }
    ]

    metric_completion = metric_client.chat.completions.create(
        model=aoai_model_name_metrics,
        messages=messages,
        temperature=0,
    )

    return metric_completion.choices[0].message.content

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 13, Finished, Available, Finished)

def get_relevance_metric(context, question, answer):    
    relevance_prompt_template = """
    Relevance measures how well the answer addresses the main aspects of the question, based on the context. Consider whether all and only the important aspects are contained in the answer when evaluating relevance. Given the context and question, score the relevance of the answer between one to five stars using the following rating scale:
    One star: the answer completely lacks relevance
    Two stars: the answer mostly lacks relevance
    Three stars: the answer is partially relevant
    Four stars: the answer is mostly relevant
    Five stars: the answer has perfect relevance

    This rating value should always be an integer between 1 and 5. So the rating produced should be 1 or 2 or 3 or 4 or 5.

    context: Marie Curie was a Polish-born physicist and chemist who pioneered research on radioactivity and was the first woman to win a Nobel Prize.
    question: What field did Marie Curie excel in?
    answer: Marie Curie was a renowned painter who focused mainly on impressionist styles and techniques.
    stars: 1

    context: The Beatles were an English rock band formed in Liverpool in 1960, and they are widely regarded as the most influential music band in history.
    question: Where were The Beatles formed?
    answer: The band The Beatles began their journey in London, England, and they changed the history of music.
    stars: 2

    context: The recent Mars rover, Perseverance, was launched in 2020 with the main goal of searching for signs of ancient life on Mars. The rover also carries an experiment called MOXIE, which aims to generate oxygen from the Martian atmosphere.
    question: What are the main goals of Perseverance Mars rover mission?
    answer: The Perseverance Mars rover mission focuses on searching for signs of ancient life on Mars.
    stars: 3

    context: The Mediterranean diet is a commonly recommended dietary plan that emphasizes fruits, vegetables, whole grains, legumes, lean proteins, and healthy fats. Studies have shown that it offers numerous health benefits, including a reduced risk of heart disease and improved cognitive health.
    question: What are the main components of the Mediterranean diet?
    answer: The Mediterranean diet primarily consists of fruits, vegetables, whole grains, and legumes.
    stars: 4

    context: The Queen's Royal Castle is a well-known tourist attraction in the United Kingdom. It spans over 500 acres and contains extensive gardens and parks. The castle was built in the 15th century and has been home to generations of royalty.
    question: What are the main attractions of the Queen's Royal Castle?
    answer: The main attractions of the Queen's Royal Castle are its expansive 500-acre grounds, extensive gardens, parks, and the historical castle itself, which dates back to the 15th century and has housed generations of royalty.
    stars: 5

    Don't explain the reasoning. The answer should include only a number: 1, 2, 3, 4, or 5.

    context: {context}
    question: {question}
    answer: {answer}
    stars:
    """

    metric_client = openai.AzureOpenAI(
        api_version=aoai_api_version,
        azure_endpoint=aoai_endpoint,
        api_key=aoai_key,
    )


    messages = [
        {
            "role": "system",
            "content": "You are an AI assistant. You are given the definition of an evaluation metric for assessing the quality of an answer in a question-answering task. Compute an accurate evaluation score using the provided evaluation metric."
        }, 
        {
            "role": "user",
            "content": relevance_prompt_template.format(context=context, question=question, answer=answer)
        }
    ]

    metric_completion = metric_client.chat.completions.create(
        model=aoai_model_name_metrics,
        messages=messages,
        temperature=0,
    )

    return metric_completion.choices[0].message.content

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 14, Finished, Available, Finished)

def get_similarity_metric(question, ground_truth, answer):
    similarity_prompt_template = """
    Equivalence, as a metric, measures the similarity between the predicted answer and the correct answer. If the information and content in the predicted answer is similar or equivalent to the correct answer, then the value of the Equivalence metric should be high, else it should be low. Given the question, correct answer, and predicted answer, determine the value of Equivalence metric using the following rating scale:
    One star: the predicted answer is not at all similar to the correct answer
    Two stars: the predicted answer is mostly not similar to the correct answer
    Three stars: the predicted answer is somewhat similar to the correct answer
    Four stars: the predicted answer is mostly similar to the correct answer
    Five stars: the predicted answer is completely similar to the correct answer

    This rating value should always be an integer between 1 and 5. So the rating produced should be 1 or 2 or 3 or 4 or 5.

    The examples below show the Equivalence score for a question, a correct answer, and a predicted answer.

    question: What is the role of ribosomes?
    correct answer: Ribosomes are cellular structures responsible for protein synthesis. They interpret the genetic information carried by messenger RNA (mRNA) and use it to assemble amino acids into proteins.
    predicted answer: Ribosomes participate in carbohydrate breakdown by removing nutrients from complex sugar molecules.
    stars: 1

    question: Why did the Titanic sink?
    correct answer: The Titanic sank after it struck an iceberg during its maiden voyage in 1912. The impact caused the ship's hull to breach, allowing water to flood into the vessel. The ship's design, lifeboat shortage, and lack of timely rescue efforts contributed to the tragic loss of life.
    predicted answer: The sinking of the Titanic was a result of a large iceberg collision. This caused the ship to take on water and eventually sink, leading to the death of many passengers due to a shortage of lifeboats and insufficient rescue attempts.
    stars: 2

    question: What causes seasons on Earth?
    correct answer: Seasons on Earth are caused by the tilt of the Earth's axis and its revolution around the Sun. As the Earth orbits the Sun, the tilt causes different parts of the planet to receive varying amounts of sunlight, resulting in changes in temperature and weather patterns.
    predicted answer: Seasons occur because of the Earth's rotation and its elliptical orbit around the Sun. The tilt of the Earth's axis causes regions to be subjected to different sunlight intensities, which leads to temperature fluctuations and alternating weather conditions.
    stars: 3

    question: How does photosynthesis work?
    correct answer: Photosynthesis is a process by which green plants and some other organisms convert light energy into chemical energy. This occurs as light is absorbed by chlorophyll molecules, and then carbon dioxide and water are converted into glucose and oxygen through a series of reactions.
    predicted answer: In photosynthesis, sunlight is transformed into nutrients by plants and certain microorganisms. Light is captured by chlorophyll molecules, followed by the conversion of carbon dioxide and water into sugar and oxygen through multiple reactions.
    stars: 4

    question: What are the health benefits of regular exercise?
    correct answer: Regular exercise can help maintain a healthy weight, increase muscle and bone strength, and reduce the risk of chronic diseases. It also promotes mental well-being by reducing stress and improving overall mood.
    predicted answer: Routine physical activity can contribute to maintaining ideal body weight, enhancing muscle and bone strength, and preventing chronic illnesses. In addition, it supports mental health by alleviating stress and augmenting general mood.
    stars: 5

    Don't explain the reasoning. The answer should include only a number: 1, 2, 3, 4, or 5.

    question: {question}
    correct answer:{ground_truth}
    predicted answer: {answer}
    stars:
    """
    
    metric_client = openai.AzureOpenAI(
        api_version=aoai_api_version,
        azure_endpoint=aoai_endpoint,
        api_key=aoai_key,
    )

    messages = [
        {
            "role": "system",
            "content": "You are an AI assistant. You will be given the definition of an evaluation metric for assessing the quality of an answer in a question-answering task. Your job is to compute an accurate evaluation score using the provided evaluation metric."
        }, 
        {
            "role": "user",
            "content": similarity_prompt_template.format(question=question, ground_truth=ground_truth, answer=answer)
        }
    ]

    metric_completion = metric_client.chat.completions.create(
        model=aoai_model_name_metrics,
        messages=messages,
        temperature=0,
    )

    return metric_completion.choices[0].message.content

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 15, Finished, Available, Finished)

Test the relevance metric:

get_relevance_metric(retrieved_context, question, answer)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 16, Finished, Available, Finished) '2'

A score of 5 means the answer is relevant. The following code gets the similarity metric:

get_similarity_metric(question, 'three', answer)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 17, Finished, Available, Finished) '5'

A score of 5 means the answer matches the ground truth answer curated by a human expert. AI-assisted metric scores can fluctuate with the same input. They're faster than using human judges.

Evaluate RAG performance on benchmark Q&A

Create function wrappers to run at scale. Wrap each function that ends with _udf (short for user-defined function) so it conforms to Spark requirements (@udf(returnType=StructType([ ... ]))) and runs calculations on large data faster across the cluster.

# UDF wrappers for RAG components
@udf(returnType=StructType([  
    StructField("retrieved_context", StringType(), True),  
    StructField("retrieved_sources", ArrayType(StringType()), True)  
]))
def get_context_source_udf(question, topN=3):
    return get_context_source(question, topN)

@udf(returnType=StringType())
def get_answer_udf(question, context):
    return get_answer(question, context)


# UDF wrapper for retrieval score
@udf(returnType=StringType())
def get_retrieval_score_udf(target_source, retrieved_sources):
    return get_retrieval_score(target_source, retrieved_sources)


# UDF wrappers for AI-assisted metrics
@udf(returnType=StringType())
def get_groundedness_metric_udf(context, answer):
    return get_groundedness_metric(context, answer)

@udf(returnType=StringType())
def get_relevance_metric_udf(context, question, answer): 
    return get_relevance_metric(context, question, answer)

@udf(returnType=StringType())
def get_similarity_metric_udf(question, ground_truth, answer):
    return get_similarity_metric(question, ground_truth, answer)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 18, Finished, Available, Finished)

Check-in #1: performance of the retriever

The following code creates the result and retrieval_score columns in the benchmark DataFrame. These columns include the RAG-generated answer and an indicator of whether the context provided to the LLM includes the article the question is based on.

df = df.withColumn("result", get_context_source_udf(df.Question)).select(df.columns+["result.*"])
df = df.withColumn('retrieval_score', get_retrieval_score_udf(df.ExtractedPath, df.retrieved_sources))
print("Aggregate Retrieval score: {:.2f}%".format((df.where(df["retrieval_score"] == 1).count() / df.count()) * 100))
display(df.select(["question", "retrieval_score",  "ExtractedPath", "retrieved_sources"]))

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 19, Finished, Available, Finished) Aggregate Retrieval score: 100.00% SynapseWidget(Synapse.DataFrame, 14efe386-836a-4765-bd88-b121f32c7cfc)

For all questions, the retriever fetches the correct context, and in most cases it's the first entry. Azure AI Search performs well. You might wonder why, in some cases, the context has two or three identical values. That's not an error - it means the retriever fetches fragments of the same article that don't fit into one chunk during splitting.

Check-in #2: performance of the response generator

Pass the question and context to the LLM to generate an answer. Store it in the generated_answer column in the DataFrame:

df = df.withColumn('generated_answer', get_answer_udf(df.Question, df.retrieved_context))

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 20, Finished, Available, Finished)

Use the generated answer, ground-truth answer, question, and context to calculate metrics. Display evaluation results for each question-answer pair:

df = df.withColumn('gpt_groundedness', get_groundedness_metric_udf(df.retrieved_context, df.generated_answer))
df = df.withColumn('gpt_relevance', get_relevance_metric_udf(df.retrieved_context, df.Question, df.generated_answer))
df = df.withColumn('gpt_similarity', get_similarity_metric_udf(df.Question, df.Answer, df.generated_answer))
display(df.select(["question", "answer", "generated_answer", "retrieval_score", "gpt_groundedness","gpt_relevance", "gpt_similarity"]))

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 21, Finished, Available, Finished) SynapseWidget(Synapse.DataFrame, 22b97d27-91e1-40f3-b888-3a3399de9d6b)

What do these values show? To make them easier to interpret, plot histograms of groundedness, relevance, and similarity. The LLM is more verbose than human ground-truth answers, which lowers the similarity metric - about half the answers are semantically correct but receive four stars as mostly similar. Most values for all three metrics are 4 or 5, which suggests RAG performance is good. There are a few outliers - for example, for the question How many species of otter are there?, the model generated There are 13 species of otter, which is correct with high relevance and similarity (5). For some reason, GPT considered it poorly grounded in the provided context and gave it one star. In the other three cases with at least one AI-assisted metric of one star, the low score points to a bad answer. The LLM occasionally mis-scores but usually scores accurately.

# Convert Spark DataFrame to Pandas DataFrame
pandas_df = df.toPandas()

selected_columns = ['gpt_groundedness', 'gpt_relevance', 'gpt_similarity']
trimmed_df = pandas_df[selected_columns].astype(int)

# Define a function to plot histograms for the specified columns
def plot_histograms(dataframe, columns):
    # Set up the figure size and subplots
    plt.figure(figsize=(15, 5))
    for i, column in enumerate(columns, 1):
        plt.subplot(1, len(columns), i)
        # Filter the dataframe to only include rows with values 1, 2, 3, 4, 5
        filtered_df = dataframe[dataframe[column].isin([1, 2, 3, 4, 5])]
        filtered_df[column].hist(bins=range(1, 7), align='left', rwidth=0.8)
        plt.title(f'Histogram of {column}')
        plt.xlabel('Values')
        plt.ylabel('Frequency')
        plt.xticks(range(1, 6))
        plt.yticks(range(0, 20, 2))


# Call the function to plot histograms for the specified columns
plot_histograms(trimmed_df, selected_columns)

# Show the plots
plt.tight_layout()
plt.show()

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 24, Finished, Available, Finished)

As a final step, save the benchmark results to a table in your lakehouse. This step is optional but highly recommended - it makes your findings more useful. When you change something in the RAG (for example, modify the prompt, update the index, or use a different GPT model in the response generator), measure the impact, quantify improvements, and detect regressions.

# create name of experiment that is easy to refer to
friendly_name_of_experiment = "rag_tutorial_experiment_1"

# Note the current date and time  
time_of_experiment = current_timestamp()

# Generate a unique GUID for all rows
experiment_id = str(uuid.uuid4())

# Add two new columns to the Spark DataFrame
updated_df = df.withColumn("execution_time", time_of_experiment) \
                        .withColumn("experiment_id", lit(experiment_id)) \
                        .withColumn("experiment_friendly_name", lit(friendly_name_of_experiment))

# Store the updated DataFrame in the default lakehouse as a table named 'rag_experiment_runs'
table_name = "rag_experiment_run_demo1" 
updated_df.write.format("parquet").mode("append").saveAsTable(table_name)

Cell output: StatementMeta(, 21cb8cd3-7742-4c1f-8339-265e2846df1d, 28, Finished, Available, Finished)

Return to the experiment results anytime to review them, compare with new experiments, and choose the configuration that works best for production.

Summary

Use AI-assisted metrics and top-N retrieval rate to build your retrieval-augmented generation (RAG) solution.

Feedback

Was this page helpful?

Last updated on 2025-10-01