NCA-GENM Question 84
Single answerYou are assisting in the development of a multimodal AI model designed to process both text and images. During testing, you notice the model performs well on individual modalities (text-only and image-only inputs) but struggles with combined multimodal inputs. Which approach is most likely to improve the model's performance on multimodal inputs?
- A
Fine-tune the model using a dataset that contains combined text and image inputs.
- B
Increase the number of layers in the model's architecture to improve its capacity.
- C
Replace the multimodal model with two separate models, one for text and one for images.
- D
Use a larger batch size during training to stabilize the gradients.
Show answer and explanation
Correct answer: A
Explanation
When a multimodal AI model struggles with combined inputs, it is often due to insufficient exposure to such inputs during training. Fine-tuning the model on a dataset containing combined text and image inputs ensures the model learns to integrate and process both modalities effectively, making this the most appropriate solution.
- A. Correct.
Fine-tuning the model with a dataset that contains combined text and image inputs allows the model to learn the relationships between the two modalities, directly addressing the performance issue with multimodal inputs.
- B. Incorrect.
Increasing the number of layers in the model's architecture may improve its capacity, but it does not specifically address the issue of poor performance on combined multimodal inputs.
- C. Incorrect.
Replacing the multimodal model with two separate models eliminates the ability to process multimodal inputs altogether, which defeats the purpose of developing a multimodal AI model.
- D. Incorrect.
Using a larger batch size during training can stabilize gradients, but it does not inherently address the issue of multimodal input performance.