Open Source AI Quiz

Answers and explanations appear as you take the quiz.

1. An open-source model can be downloaded without a fee. What follows about the cost of operating it?
   • Running, maintaining and evaluating it can still require substantial resources
   • The license guarantees free hosting
   • No hardware is needed once weights are public
   • Evaluation becomes unnecessary because the release is open
2. A team is evaluating one release. Which statement about component licenses matches the guide?
   • Every component automatically shares one license
   • Code, weights, datasets, and supporting assets may have different terms
   • Only the dataset can have restrictions
   • A model card overrides all licenses
3. What does the guide say openness can support without guaranteeing?
   • Commercial use without checking terms
   • Automatic reproducibility in every environment
   • Inspection, adaptation, and research, while reproducibility and quality are not guaranteed
   • Identical results on all hardware
4. A team cannot reproduce a published result from a released model. Which barrier is listed in the guide?
   • Documented preprocessing steps
   • A complete configuration matching the experiment
   • A clearly stated license
   • Missing configuration or undocumented preprocessing
5. Before adopting a release, which set of facts does the guide recommend recording?
   • Only its download count
   • Exact version, licenses, provenance, and deployment requirements
   • Only the organization that announced it
   • Only the result of one hosted demonstration
6. According to the guide, what does OSI’s preferred form for modification address?
   • Data information, code, and parameters
   • Every raw training record made public
   • Only the hosted web demo
   • Only the model’s marketing name
7. A release offers weights but omits training code and restricts some uses. Which description best follows the guide?
   • Call it fully reproducible
   • Call it unrestricted open source because weights are available
   • Record those limits and compare the release with the stated definition
   • Assume its missing parts are public
8. After making a model publicly available, what responsibility remains with the organization operating it?
   • Avoiding all version records
   • None, because public availability transfers responsibility
   • Only paying for hosted inference
   • Testing for the intended task and maintaining security and privacy controls