Module 10 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. A good visual prompt…
- A. is open-ended
- B. defines what a correct answer looks like, and how to say 'cannot tell'
- C. is as short as possible
- D. asks for a description
2. Asked about something not visible in the image, the model will usually…
- A. refuse
- B. produce a confident, plausible answer
- C. return an error
- D. ask for clarification
3. Which is a known weak point answered fluently?
- A. identifying a large object
- B. counting past about four objects
- C. naming the dominant colour
- D. detecting presence
4. Requiring a region with every visual claim…
- A. slows the model down
- B. makes the answer checkable in seconds
- C. is only useful for detection
- D. reduces accuracy
5. Refusal conditions like region size and blur should be…
- A. written into the prompt
- B. computed before the model is called
- C. learned by the model
- D. checked by a reviewer
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - defines what a correct answer looks like, and how to say 'cannot tell'
Open prompts produce output you can only judge by reading, which does not scale.
2. B - produce a confident, plausible answer
A plausible completion about an unobservable property is indistinguishable in form from an observation.
3. B - counting past about four objects
Small-text reading and spatial relations are the other two, and all three produce grammatical confident answers.
4. B - makes the answer checkable in seconds
Validate the region against the image bounds - a model asked for coordinates will produce them, including impossible ones.
5. B - computed before the model is called
They are measurable for free, and the refusal then costs a subtraction rather than a model call.
Practical Check
Apply this module to your own work: complete the module activity for *Multimodal Prompts and Vision-Language Model Limits*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.