Vision-language models (GPT-4V) in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the core capability of GPT-4V

What is the primary advantage of GPT-4V compared to traditional language-only models?

A It focuses solely on generating images from text prompts.

B It replaces the need for any text input by using images alone.

C It can process and understand both images and text inputs simultaneously.

D It only improves text generation speed without image understanding.

💡 Hint

Think about what 'vision-language' means in the model's name.

GPT-4V combines visual and textual understanding, allowing it to interpret images and text together, unlike traditional models that only handle text.

❓ Model Choice

intermediate

Choosing the right model for multimodal tasks

You want to build an application that answers questions about photos users upload. Which model is best suited for this task?

A GPT-3, because it is the most advanced text-only model.

B GPT-4V, because it can understand images and text together.

C A convolutional neural network (CNN) trained only on image classification.

D A text summarization model trained on news articles.

💡 Hint

Consider which model can handle both images and text inputs.

GPT-4V is designed to understand images and text jointly, making it ideal for answering questions about photos, unlike text-only or image-only models.

❓ Predict Output

advanced

Predicting GPT-4V output for image-text input

Given an image of a red apple and the text prompt 'What fruit is this?', what is the most likely output from GPT-4V?

A "This is a red apple."

B "This is a green apple."

C "This is a banana."

D "I cannot see any fruit in the image."

💡 Hint

GPT-4V identifies objects in images and answers questions about them.

GPT-4V uses its vision-language ability to recognize the red apple in the image and respond accurately to the question.

❓ Metrics

advanced

Evaluating GPT-4V's performance on multimodal tasks

Which metric best measures GPT-4V's accuracy in answering questions about images?

A Perplexity, which measures how well the model predicts the next word in text.

B BLEU score, which measures similarity between generated and reference text only.

C Image classification accuracy, which only measures image label correctness without text.

D Multimodal accuracy, which checks if the model's answer matches the correct text for the given image and question.

💡 Hint

Think about a metric that evaluates both image understanding and text generation.

Multimodal accuracy evaluates if the model correctly answers questions based on both image content and text input, which is essential for GPT-4V.

🔧 Debug

expert

Diagnosing GPT-4V's incorrect image question answering

You notice GPT-4V often answers incorrectly when images contain multiple objects. What is the most likely cause?

A The model's attention mechanism may not effectively focus on the relevant object in complex scenes.

B The model cannot process images larger than 64x64 pixels.

C The text input is ignored when multiple objects are present in the image.

D The model only works with black and white images, so color images cause errors.

💡 Hint

Consider how the model handles complex visual scenes with many details.

GPT-4V uses attention to link text and image parts; if it struggles to focus on the right object among many, answers can be wrong.

Practice

1. What is the main capability of vision-language models like GPT-4V?

easy

A. They understand and generate responses based on both images and text.

B. They only process text data without images.

C. They only analyze images without any text understanding.

D. They translate languages without any image input.

Solution

1. Step 1: Understand the model's input types

   Vision-language models take both images and text as input to understand context.

2. Step 2: Recognize the model's output capabilities

   They generate responses that relate to both the visual content and the text prompt.

3. Final Answer:

   They understand and generate responses based on both images and text. -> Option A

4. Quick Check:

   Vision + Language = Both inputs [OK]

Hint: Vision-language means both image and text understanding [OK]

Common Mistakes:

• Thinking the model only works with text
• Assuming it only processes images
• Confusing translation with vision-language tasks

2. Which of the following is the correct way to prompt GPT-4V to describe an image?

easy

A. Translate the text: [image]

B. Describe the image: [image]

C. Calculate the sum: [image]

D. Play music from: [image]

Solution

1. Step 1: Identify the prompt that asks for image description

   Only Describe the image: [image] clearly requests a description of the image content.

2. Step 2: Eliminate unrelated commands

   Options B, C, and D ask for translation, calculation, or music playing, which are unrelated to image description.

3. Final Answer:

   <code>Describe the image: [image]</code> -> Option B

4. Quick Check:

   Prompt matches task: describe image [OK]

Hint: Look for prompt asking to describe the image [OK]

Common Mistakes:

• Choosing prompts unrelated to images
• Confusing translation with description
• Ignoring the image context in the prompt

3. Given the following code snippet using GPT-4V API, what will be the output?

response = gpt4v.ask(image='cat.jpg', prompt='What animal is in the picture?')
print(response)

medium

A. SyntaxError: missing argument

B. "I cannot see any animal in the picture."

C. "The animal in the picture is a cat."

D. "The picture shows a dog."

Solution

1. Step 1: Understand the prompt and image input

   The prompt asks what animal is in the image named 'cat.jpg', which likely contains a cat.

2. Step 2: Predict the model's response

   GPT-4V will analyze the image and respond with the correct animal, which is a cat.

3. Final Answer:

   "The animal in the picture is a cat." -> Option C

4. Quick Check:

   Image name + prompt = cat answer [OK]

Hint: Match image content with prompt question [OK]

Common Mistakes:

• Assuming the model cannot see images
• Expecting error due to missing arguments
• Confusing animal types in output

4. Identify the error in this GPT-4V usage code snippet:

response = gpt4v.ask(prompt='Describe this image.')
print(response)

medium

A. Missing image input argument in the ask function.

B. The prompt text is too short.

C. The print statement is incorrect syntax.

D. The ask function does not exist in GPT-4V.

Solution

1. Step 1: Check required inputs for vision-language query

   GPT-4V requires both an image and a prompt to answer about the image.

2. Step 2: Identify missing argument

   The code only provides a prompt but no image, which is necessary for vision understanding.

3. Final Answer:

   Missing image input argument in the ask function. -> Option A

4. Quick Check:

   Image missing in ask() call [OK]

Hint: Vision queries need both image and prompt [OK]

Common Mistakes:

• Ignoring the need for image input
• Thinking prompt length causes error
• Assuming print syntax is wrong

5. You want GPT-4V to find all objects in a complex image and list them with counts. Which approach is best?

hard

A. Send multiple images without prompts and combine answers manually.

B. Send only the image without any prompt and expect a list.

C. Use a prompt asking to translate the image content to another language.

D. Use a prompt like List all objects and their counts in this image: [image] and parse the response.

Solution

1. Step 1: Understand the task requirements

   The task is to identify and count objects in one image, so a clear prompt is needed.

2. Step 2: Choose the prompt that requests object listing and counting

   Use a prompt like List all objects and their counts in this image: [image] and parse the response, which explicitly asks for listing objects and counts, which GPT-4V can handle.

3. Step 3: Eliminate other options

   Sending only the image without any prompt lacks specific task instructions. Using a prompt to translate the image content is unrelated to object detection. Sending multiple images without prompts and combining answers manually is inefficient and unclear.

4. Final Answer:

   Use a prompt like <code>List all objects and their counts in this image: [image]</code> and parse the response. -> Option D

5. Quick Check:

   Clear prompt + image = correct object list [OK]

Hint: Always include clear prompt with image for object detection [OK]

Common Mistakes:

• Sending image without prompt expecting detailed output
• Confusing translation with object detection
• Using multiple images without clear instructions