Test cases for tool-using agents in Agentic AI - Model Metrics & Evaluation

Metrics & Evaluation - Test cases for tool-using agents

Which metric matters for Test cases for tool-using agents and WHY

For tool-using agents, key metrics include task success rate and tool invocation accuracy. Task success rate shows how often the agent completes the goal correctly using tools. Tool invocation accuracy measures if the agent calls the right tool at the right time. These metrics matter because the agent must both choose and use tools properly to solve problems.

Confusion matrix or equivalent visualization

      | Predicted Tool Use |
      |-------------------|
      | True Positive (TP): Agent correctly uses the tool when needed
      | False Positive (FP): Agent uses tool when not needed
      | True Negative (TN): Agent correctly does not use tool when not needed
      | False Negative (FN): Agent fails to use tool when needed

      Total samples = TP + FP + TN + FN
    

Precision vs Recall tradeoff with examples

Precision means when the agent uses a tool, it is usually the right choice. High precision avoids wasting resources on wrong tools.

Recall means the agent uses the tool whenever it is needed. High recall avoids missing important tool uses.

Example: For a cooking assistant agent, high precision means it rarely uses the blender when not needed. High recall means it always uses the blender when the recipe calls for it.

What good vs bad metric values look like

Good: Task success rate above 90%, tool invocation precision and recall above 85%. This means the agent reliably uses tools correctly and completes tasks.

Bad: Task success rate below 60%, precision or recall below 50%. The agent often misuses tools or misses using them, leading to failed tasks.

Common pitfalls in metrics

• Accuracy paradox: High overall accuracy can hide poor tool use if most tasks don't require tools.
• Data leakage: Testing on tasks seen during training inflates success rates.
• Overfitting: Agent memorizes tool use patterns but fails on new tasks.
• Ignoring timing: Using the right tool too late can still cause task failure but may not be captured by simple metrics.

Self-check question

Your tool-using agent has 98% task success rate but only 12% recall on tool invocation. Is it good for production? Why or why not?

Answer: No, it is not good. The agent rarely uses tools when needed (low recall), so it might succeed on simple tasks but fail on complex ones requiring tools. High task success alone can be misleading.

Key Result

For tool-using agents, both task success rate and tool invocation precision/recall are essential to measure correct and timely tool use.

Practice

1. What is the main purpose of writing test cases for tool-using agents?

easy

A. To add more tools to the agent

B. To make agents run faster

C. To check if agents use tools correctly and handle errors

D. To reduce the size of the agent's code

Solution

1. Step 1: Understand the role of test cases

   Test cases are designed to verify that the agent behaves as expected, especially when using tools.

2. Step 2: Identify the main goal for tool-using agents

   For agents that use tools, tests ensure they use these tools correctly and handle any errors gracefully.

3. Final Answer:

   To check if agents use tools correctly and handle errors -> Option C

4. Quick Check:

   Test cases purpose = check tool use and errors [OK]

Hint: Test cases verify correct tool use and error handling [OK]

Common Mistakes:

• Thinking test cases speed up agents
• Believing test cases reduce code size
• Assuming test cases add tools

2. Which of the following is the correct way to write a test case for a tool-using agent in Python?

easy

A. test agent tool: assert agent.use_tool('calculator', '2+2') == 4

B. def test_agent_tool(): assert agent.use_tool('calculator', '2+2') == 4

C. def test_agent_tool: assert agent.use_tool('calculator', '2+2') == 4

D. def test_agent_tool() assert agent.use_tool('calculator', '2+2') == 4

Solution

1. Step 1: Check Python function syntax

   Python test functions start with 'def', have parentheses, and a colon at the end.

2. Step 2: Verify assertion syntax

   The assert statement must be inside the function and correctly compare expected output.

3. Final Answer:

   def test_agent_tool(): assert agent.use_tool('calculator', '2+2') == 4 -> Option B

4. Quick Check:

   Correct Python test function syntax = def test_agent_tool(): assert agent.use_tool('calculator', '2+2') == 4 [OK]

Hint: Remember Python functions need parentheses and colon [OK]

Common Mistakes:

• Omitting parentheses in function definition
• Missing colon after function header
• Incorrect assert statement placement

3. Given this test case code snippet, what will be the output if the agent returns 5 instead of 4?

def test_agent_tool():
    result = agent.use_tool('calculator', '2+2')
    assert result == 4
    print('Test passed')

medium

A. Test passed

B. SyntaxError

C. No output

D. AssertionError

Solution

1. Step 1: Understand assert behavior

   If the assert condition is false, Python raises an AssertionError and stops execution.

2. Step 2: Check the test condition

   The test expects result == 4, but agent returns 5, so assert fails.

3. Final Answer:

   AssertionError -> Option D

4. Quick Check:

   Assert fails if values differ = AssertionError [OK]

Hint: Assert fails if expected and actual differ [OK]

Common Mistakes:

• Thinking print runs after failed assert
• Confusing AssertionError with SyntaxError
• Assuming no output on failure

4. Identify the error in this test case for a tool-using agent:

def test_agent_tool():
    result = agent.use_tool('search', 'weather today')
    assert result = 'sunny'
    print('Test passed')

medium

A. Using '=' instead of '==' in assert

B. Missing parentheses in print

C. Wrong function name

D. Agent tool name is invalid

Solution

1. Step 1: Check assert statement syntax

   In Python, '=' is for assignment, '==' is for comparison. Assert needs '==' to compare values.

2. Step 2: Verify other parts

   Print has parentheses, function name is valid, and tool name is plausible.

3. Final Answer:

   Using '=' instead of '==' in assert -> Option A

4. Quick Check:

   Assert needs '==' for comparison [OK]

Hint: Assert compares with '==', not '=' [OK]

Common Mistakes:

• Confusing assignment '=' with comparison '=='
• Ignoring syntax errors in assert
• Assuming print needs no parentheses

5. You want to test an agent that uses a calculator tool to handle multiple expressions. Which test case best checks if the agent correctly handles both valid and invalid inputs?

hard

A. def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', 'abc') == 'error'

B. def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', '3/0') == 0

C. def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', '') == ''

D. def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', null) == null"

Solution

1. Step 1: Check valid input test

   All options test '3*3' == 9 correctly, which is good for valid input.

2. Step 2: Check invalid input handling

   def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', 'abc') == 'error' expects 'abc' input to return 'error', which correctly tests error handling. Others expect incorrect or unclear outputs.

3. Final Answer:

   def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', 'abc') == 'error' -> Option A

4. Quick Check:

   Test valid and invalid inputs properly = def test_calc(): assert agent.use_tool('calculator', '3*3') == 9; assert agent.use_tool('calculator', 'abc') == 'error' [OK]

Hint: Test both valid and invalid inputs explicitly [OK]

Common Mistakes:

• Expecting wrong output for invalid input
• Not testing error cases
• Assuming empty or null inputs return themselves