최신 NVIDIA-Certified Associate NCA-GENM 무료샘플문제:

1. You're building a system to translate customer service chat logs into summaries that a human agent can quickly review The chat logs are often informal, contain slang, and have grammatical errors. Which prompt engineering technique is MOST likely to improve the quality and accuracy of the summaries generated by a large language model (LLM)?

A) Using a few-shot prompt with several examples of chat logs and their ideal summaries, explicitly demonstrating how to handle informality and errors.
B) Using a negative constraint prompt, explicitly stating what the LLM should not include in the summary (e.g., 'Do not include greetings or farewells.').
C) Using a template prompt with predefined sections and keywords to guide the summarization process and ensure consistency across different chat logs.
D) Using chain-of-thought prompting to encourage the LLM to explain its reasoning process before generating the summary.
E) Using a zero-shot prompt with a simple instruction like 'Summarize this chat log.'

2. You are experimenting with different multimodal transformer architectures for a video understanding task. You are using a large pre- trained model and fine-tuning it on your specific dataset. You observe that the model is overfitting and struggling to generalize to unseen videos. Which of the following techniques would be most effective in mitigating overfitting in this scenario? (Choose two)

A) Implement weight decay and dropout regularization.
B) Reduce the number of transformer layers in the model.
C) Use a smaller pre-trained model.
D) Increase the batch size significantly.
E) Employ data augmentation techniques specifically designed for video data (e.g., temporal jittering, random cropping).

3. You are training a multimodal model with text and audio inputs. You notice that the audio modality dominates the training process, and the text modality is not contributing significantly to the final performance. Which of the following strategies can you use to address this modality imbalance? (Select TWO)

A) Increase the size of the audio dataset.
B) Increase the learning rate for the text encoder.
C) Remove the audio modality altogether to force the model to rely on text
D) Apply a modality-specific weighting scheme to the loss function, giving more weight to the text loss
E) Decrease the batch size for the audio data

4. You are tasked with building a system that generates realistic images based on both textual descriptions and a semantic segmentation map. The segmentation map provides spatial information about the objects present in the scene. Which of the following generative architectures is MOST appropriate for this multimodal task?

A) Conditional Generative Adversarial Network (cGAN) with both text and segmentation map as conditions.
B) Variational Autoencoder (VAE)
C) Vanilla Generative Adversarial Network (GAN)
D) Autoregressive model like PixelCNN
E) Diffusion model without conditioning

5. You are developing a multimodal sentiment analysis model that combines text reviews and product images. You observe that the model's performance is significantly better when only text is used, compared to when both text and images are combined. What are the potential reasons for this performance degradation, and how can you address them effectively? (Choose two)

A) The image features are noisy or of poor quality, confusing the model.
B) The model is not properly aligning the text and image features, leading to conflicting signals.
C) The image features are irrelevant to the sentiment expressed in the text.
D) The text encoder is too complex, hindering the model's ability to process image information.
E) The model is overfitting to the image features.

질문과 대답: