Advanced level · Module 5: Responsible Deep Learning · Lesson 4 of 4
Documentation and Red-Teaming
Document deep learning systems and test them deliberately for failures and misuse.
In this lesson you will
- Write model cards for deep learning systems
- Conduct a structured red-team exercise
Key concepts
model cardintended and prohibited usesred-teamingfailure documentation
The lesson
A model card documents a model's purpose, training data, performance across groups and conditions, limits, prohibited uses and contacts.
Red-teaming means deliberately trying to make a system fail or be misused: unusual inputs, adversarial images, offensive text, attempts to extract private data.
Record failures and fix them or add safeguards before deployment; update the model card with what was found.
Step by step
- Write model cards for the two provided models.
- Plan red-team tests.
- Run the tests and record failures.
- Update the model cards.
Practice activity
Complete the model cards and red-team report from the practical activity.
Expected result: Complete model cards and a red-team report with safeguards.
Check your understanding
-
What is red-teaming?
Show answer
Deliberately trying to make a system fail or be misused, to find weaknesses.
-
What should a model card's prohibited-use section state?
Show answer
Uses the model must not be applied to.
Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.