When to fine-tune vs prompt engineer in Prompt Engineering / GenAI - Metrics Comparison

Metrics & Evaluation - When to fine-tune vs prompt engineer

Which metric matters and WHY

When deciding between fine-tuning a model or prompt engineering, key metrics to watch are task accuracy, response relevance, and latency. Fine-tuning aims to improve accuracy and relevance by changing the model's knowledge, while prompt engineering tries to get better answers without changing the model. Measuring accuracy or quality of answers helps decide which approach works best.

Confusion matrix or equivalent visualization

Task: Classify user intent from text

Confusion Matrix Example:
          Predicted
          Yes   No
Actual Yes  80   20
       No   15   85

- Fine-tuning can improve these numbers by learning from more examples.
- Prompt engineering tries to reduce errors by better question phrasing.

Precision vs Recall tradeoff with examples

Fine-tuning improves both precision and recall by teaching the model new patterns. It is good when you have many examples and want consistent, high-quality results.

Prompt engineering is faster and cheaper but may only improve precision or recall slightly. It is useful when you want quick fixes or have limited data.

Example: For a customer support bot, fine-tuning can reduce missed questions (higher recall). Prompt engineering can help avoid wrong answers (higher precision) by clearer prompts.

What "good" vs "bad" metric values look like

Good: Accuracy above 85%, balanced precision and recall, fast response time.

Bad: Accuracy below 60%, very low recall (missing many correct answers), or very low precision (many wrong answers).

If prompt engineering cannot reach good metrics, fine-tuning is needed.

Common pitfalls in metrics

• Accuracy paradox: High accuracy can be misleading if data is imbalanced.
• Overfitting: Fine-tuned models may perform well on training data but poorly on new data.
• Data leakage: Using test data during fine-tuning inflates metrics falsely.
• Ignoring latency: Fine-tuning can increase response time, hurting user experience.
• Prompt bias: Poor prompt design can hide model weaknesses.

Self-check question

Your chatbot has 98% accuracy but only 12% recall on urgent requests. Is it good for production? Why or why not?

Answer: No, because it misses most urgent requests (low recall). This can cause serious problems. You should improve recall, possibly by fine-tuning or better prompt engineering.

Key Result

Fine-tuning improves accuracy and recall by changing the model, while prompt engineering tweaks inputs for quick gains; choose based on data, cost, and desired quality.

Practice

1. What is the main difference between fine-tuning a model and prompt engineering?

easy

A. Fine-tuning is faster than prompt engineering.

B. Fine-tuning changes the prompt format, while prompt engineering changes the model's weights.

C. Fine-tuning changes the model's knowledge, while prompt engineering changes how you ask questions.

D. Prompt engineering requires retraining the model.

Solution

1. Step 1: Understand fine-tuning

   Fine-tuning means adjusting the model's internal settings (weights) to better fit specific data or tasks.

2. Step 2: Understand prompt engineering

   Prompt engineering means changing the way you ask the model questions without changing the model itself.

3. Final Answer:

   Fine-tuning changes the model's knowledge, while prompt engineering changes how you ask questions. -> Option C

4. Quick Check:

   Fine-tune = model change, prompt engineer = question change [OK]

Hint: Fine-tune = model change; prompt engineer = question change [OK]

Common Mistakes:

• Confusing prompt engineering with model retraining
• Thinking fine-tuning only changes prompts
• Believing prompt engineering is slower than fine-tuning

2. Which of the following is a correct way to describe prompt engineering?

easy

A. Adjusting the input text to get better model responses.

B. Changing the model's training data to improve accuracy.

C. Rebuilding the model architecture from scratch.

D. Adding new layers to the model for customization.

Solution

1. Step 1: Identify prompt engineering meaning

   Prompt engineering means changing how you write or format the input text to guide the model's answers.

2. Step 2: Check options

   Only Adjusting the input text to get better model responses. describes adjusting input text, which matches prompt engineering.

3. Final Answer:

   Adjusting the input text to get better model responses. -> Option A

4. Quick Check:

   Prompt engineering = input text change [OK]

Hint: Prompt engineering means changing input text, not model structure [OK]

Common Mistakes:

• Mixing prompt engineering with model retraining
• Thinking prompt engineering changes model layers
• Confusing prompt engineering with data augmentation

3. Consider you want to improve a model's answers for a very specific medical dataset. Which approach will likely give better results?

medium

A. Only use prompt engineering to rewrite questions.

B. Use random prompts without changes.

C. Ignore the dataset and use the base model.

D. Fine-tune the model with the medical dataset.

Solution

1. Step 1: Understand the task

   Improving answers for a specific medical dataset requires the model to learn new, specialized knowledge.

2. Step 2: Choose the best method

   Fine-tuning the model with the medical data updates its knowledge, making it better for this task.

3. Final Answer:

   Fine-tune the model with the medical dataset. -> Option D

4. Quick Check:

   Specific data needs fine-tuning [OK]

Hint: Use fine-tuning for specialized data, not just prompt changes [OK]

Common Mistakes:

• Thinking prompt engineering alone fixes specialized knowledge
• Ignoring fine-tuning for domain-specific tasks
• Assuming base model works best without adaptation

4. You tried prompt engineering but the model still gives poor answers for your task. What is the most likely fix?

medium

A. Fine-tune the model with relevant data.

B. Change the model architecture.

C. Use shorter prompts.

D. Restart the model server.

Solution

1. Step 1: Analyze the problem

   If prompt engineering fails to improve answers, the model likely lacks task-specific knowledge.

2. Step 2: Choose the fix

   Fine-tuning with relevant data updates the model's knowledge to improve answers.

3. Final Answer:

   Fine-tune the model with relevant data. -> Option A

4. Quick Check:

   Poor answers + prompt fail = fine-tune needed [OK]

Hint: If prompts fail, fine-tune with data [OK]

Common Mistakes:

• Trying unrelated fixes like changing architecture
• Assuming shorter prompts fix knowledge gaps
• Restarting server won't improve model knowledge

5. You have a chatbot that answers general questions well but struggles with your company's product details. You want to improve it quickly without retraining. What should you do?

hard

A. Ignore product details and focus on general answers.

B. Use prompt engineering to add product info in the questions.

C. Replace the chatbot with a new model.

D. Fine-tune the entire model with product manuals.

Solution

1. Step 1: Identify constraints

   You want a quick improvement without retraining the model.

2. Step 2: Choose the best approach

   Prompt engineering lets you add product info in questions to guide the model without retraining.

3. Final Answer:

   Use prompt engineering to add product info in the questions. -> Option B

4. Quick Check:

   Quick fix without retrain = prompt engineering [OK]

Hint: Quick fix without retrain? Use prompt engineering [OK]

Common Mistakes:

• Thinking fine-tuning is always fastest
• Replacing model unnecessarily
• Ignoring product info causes poor answers