Regression testing for agent changes in Agentic AI - Model Metrics & Evaluation

Metrics & Evaluation - Regression testing for agent changes

Which metric matters for regression testing and WHY

When we update an AI agent, we want to make sure it still works well. For regression testing, we focus on error metrics like Mean Absolute Error (MAE), Mean Squared Error (MSE), or Root Mean Squared Error (RMSE). These numbers tell us how far the agent's predictions are from the true answers.

We compare these errors before and after changes. If errors get bigger, the new agent might have problems. So, these metrics help us catch mistakes introduced by updates.

Confusion matrix or equivalent visualization

Regression tasks don't use confusion matrices because they predict numbers, not categories. Instead, we look at error values.

Example errors before and after agent update:

| Metric | Before Update | After Update |
|--------|---------------|--------------|
| MAE    | 2.5           | 3.8          |
| MSE    | 9.0           | 15.0         |
| RMSE   | 3.0           | 3.87         |

Higher errors after update mean the agent's predictions got worse.
    

Tradeoff: Stability vs Improvement

When changing an agent, we want it to improve but also stay stable. If the agent's error gets smaller, that's good. But if it gets bigger, it means the update caused problems.

Sometimes, a small increase in error is okay if the agent gains new skills. But big error jumps mean we should fix the update.

Think of it like fixing a car: you want it to run better, not worse after repairs.

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

Good: After agent changes, error metrics stay the same or get smaller. For example, MAE stays around 2.5 or drops to 2.0.

Bad: Errors increase a lot, like MAE jumping from 2.5 to 5.0. This means the agent's predictions are less accurate.

Good regression testing means catching these bad changes before releasing the agent.

Common pitfalls in regression testing metrics

• Ignoring small error changes: Sometimes small error increases are normal, but ignoring big jumps is risky.
• Testing on different data: Comparing errors on different test sets can mislead results.
• Overfitting to test data: If the agent is tuned too much on test data, errors look good but real performance drops.
• Not tracking multiple metrics: Using only one error metric can hide problems. Check MAE, MSE, and RMSE together.

Self-check question

Your agent update shows 98% accuracy but the Mean Absolute Error increased from 2.0 to 6.0. Is this good?

Answer: No. Accuracy is not a good metric for regression. The big increase in MAE means predictions are less accurate. The update likely caused problems and needs review.

Key Result

For regression testing agent changes, tracking error metrics like MAE and MSE before and after updates is key to detect performance drops.

Practice

1. What is the main purpose of regression testing for agent changes?

easy

A. To check if new changes break old agent behavior

B. To improve the agent's speed

C. To add new features to the agent

D. To change the agent's user interface

Solution

1. Step 1: Understand regression testing goal

   Regression testing is done to ensure that recent changes do not break existing functionality.

2. Step 2: Match purpose with options

   To check if new changes break old agent behavior clearly states checking if new changes break old behavior, which matches the goal.

3. Final Answer:

   To check if new changes break old agent behavior -> Option A

4. Quick Check:

   Regression testing = check old behavior intact [OK]

Hint: Regression testing checks old features still work after changes [OK]

Common Mistakes:

• Thinking regression testing adds new features
• Confusing regression testing with performance testing
• Assuming regression testing changes UI

2. Which of the following is the correct way to define a test case for regression testing an agent in Python?

easy

A. def test_agent(): assert agent.run(input) == expected_output

B. test agent run input equals expected output

C. def test_agent: return agent.run(input) == expected_output

D. function test_agent() { return agent.run(input) == expected_output; }

Solution

1. Step 1: Identify correct Python function syntax

   Python functions start with 'def', have parentheses, and a colon.

2. Step 2: Check assertion usage

   def test_agent(): assert agent.run(input) == expected_output uses 'assert' correctly to compare output, matching Python test style.

3. Final Answer:

   def test_agent(): assert agent.run(input) == expected_output -> Option A

4. Quick Check:

   Python test function with assert = def test_agent(): assert agent.run(input) == expected_output [OK]

Hint: Python test functions start with def and use assert [OK]

Common Mistakes:

• Missing parentheses or colon in function definition
• Using non-Python syntax
• Not using assert for test checks

3. Given the code below, what will be the output of the regression test?

class Agent:
    def run(self, x):
        return x * 2

def test_agent():
    agent = Agent()
    result = agent.run(3)
    assert result == 6
    print('Test passed')

test_agent()

medium

A. SyntaxError

B. Test passed

C. AssertionError

D. No output

Solution

1. Step 1: Understand agent run method

   The method multiplies input by 2, so run(3) returns 6.

2. Step 2: Check assertion and print

   The assertion checks if result == 6, which is true, so no error occurs and 'Test passed' prints.

3. Final Answer:

   Test passed -> Option B

4. Quick Check:

   3 * 2 = 6, assertion true, prints message [OK]

Hint: Check method output matches assertion to predict test result [OK]

Common Mistakes:

• Assuming assertion fails without checking output
• Confusing syntax errors with logic errors
• Ignoring print statement after assertion

4. Identify the error in the following regression test code and select the fix:

def test_agent():
    agent = Agent()
    result = agent.run(5)
    if result = 10:
        print('Test passed')
    else:
        print('Test failed')

medium

A. Replace print with return statements

B. Add parentheses around the if condition

C. Change '=' to '==' in the if condition

D. Remove else block

Solution

1. Step 1: Identify syntax error in if condition

   The single '=' is an assignment, not a comparison, causing a syntax error.

2. Step 2: Correct the comparison operator

   Replace '=' with '==' to compare values properly in the if statement.

3. Final Answer:

   Change '=' to '==' in the if condition -> Option C

4. Quick Check:

   Use '==' for comparison in if statements [OK]

Hint: Use '==' to compare, '=' assigns values [OK]

Common Mistakes:

• Using '=' instead of '==' in conditions
• Adding unnecessary parentheses in Python if
• Thinking print must be replaced with return

5. You updated your agent's decision logic. How should you design regression tests to ensure old behaviors remain correct while testing new features?

hard

A. Test randomly without expected outputs to save time

B. Only test new features since old ones worked before

C. Remove old tests to avoid conflicts with new logic

D. Create test cases for both old expected outputs and new expected outputs

Solution

1. Step 1: Understand regression test purpose

   Regression tests verify that old behaviors still work after changes.

2. Step 2: Design tests covering old and new behaviors

   Include test cases for old expected outputs and new expected outputs to check both.

3. Final Answer:

   Create test cases for both old expected outputs and new expected outputs -> Option D

4. Quick Check:

   Test old and new outputs to ensure full correctness [OK]

Hint: Test old and new cases to catch breaks early [OK]

Common Mistakes:

• Ignoring old tests after updates
• Deleting old tests to simplify
• Skipping expected outputs in tests