Nvidia Nemotron Models Quiz

Answers and explanations appear as you take the quiz.

1. A major distinctive purpose of the Nemotron-4 340B family is:
   • Generating license-friendly synthetic training data
   • Rendering 3D graphics
   • Mining cryptocurrency
   • Compressing video files
2. Llama 3.1 Nemotron 70B is built by Nvidia on top of a base model from which company?
   • Google
   • Meta
   • OpenAI
   • Anthropic
3. What is the role of a Nemotron 'reward model'?
   • It pays developers for using GPUs
   • It generates random noise
   • It scores the quality of responses to guide training and data filtering
   • It encrypts model weights
4. What practical capability does NVIDIA’s open Nemotron release offer developers?
   • A guarantee that all hardware is free
   • A replacement for model evaluation
   • A promise that no fine-tuning is needed
   • A model and associated tools they can adapt for their own applications, subject to its license
5. Which Nvidia framework or product is commonly used to deploy Nemotron models?
   • NeMo and NIM microservices
   • Microsoft Word
   • Adobe Photoshop
   • Oracle Database
6. The technique Nvidia used to sharpen Nemotron 70B's helpfulness relied heavily on:
   • Hand-writing every answer
   • Reinforcement learning from human feedback with a custom reward model
   • Removing all training data
   • Disabling the GPU
7. In the Nemotron synthetic-data pipeline described in the guide, what does the reward model do?
   • Writes the GPU kernel for the language model
   • Converts all weights to ternary values
   • Scores candidate responses on qualities such as helpfulness and correctness
   • Assigns a class label to every image
8. Which preference prompts does NVIDIA name in the Llama-3.1-Nemotron-70B-Instruct model card?
   • ImageNet
   • Common Crawl RefinedWeb
   • GSM8K
   • HelpSteer2-Preference