A convention-center IT analyst attaches a facility photo and a text complaint to one request so the assistant can locate the issue in the image and draft a work order. What should the multimodal prompt emphasize?
Select an answer to reveal the explanation.
Short Explanation
Two clues, one job: the photo and the complaint need joint marching orders—what to look for and what to write. A multimodal prompt that ties them together beats telling the model to ignore the picture or dream up a ticket.
Full Explanation
Multimodal prompts should state how each modality contributes to the task. When a convention-center analyst attaches a facility photo and a text complaint, joint instructions tell the model to localize the issue in the image and use the complaint details to draft a coherent work order. Clear cross-modal guidance reduces the chance the model ignores one channel or invents facilities damage that the photo does not show.
Text-only instructions that forbid looking at the attached photo fail by concept: they discard the visual evidence the ticket depends on. Asking to ignore the complaint and invent a random work order fails because it severs grounding in the resident report and produces untrusted maintenance actions. Demanding to fine-tune a new vision model before answering this one ticket fails because single-ticket multimodal prompting does not require training a new model; fine-tuning is a separate program decision.
Exam caveat: Joint instructions improve use of photo plus text but do not guarantee correct localization—facility lighting, blur, or ambiguous damage still need human confirmation before dispatch. Privacy rules may restrict what building areas can be stored or sent to external vision models. On the exam, multimodal design means specifying roles for each input, not refusing a modality or jumping to fine-tuning.
Operational check: For a sample ticket, require the assistant to cite visible landmarks from the photo and quoted complaint phrases in the draft work order, then have facilities staff confirm the pin location on a floor plan before creating the CMMS record.