Тест за ограничено и управлявано от граматиката генериране

Отговорите и обясненията се появяват, докато отговаряте на теста.

1. At which stage does the described hard grammar constraint restrict a language model's continuation choices?
   • Only while collecting pretraining documents
   • Only after executing every emitted instruction
   • Decoding, before the next token is selected
   • By retraining the model weights after every user request
2. Which formal rule representations are discussed in the cited grammar-guided generation research?
   • Only image-file color palettes
   • Only a spreadsheet of model parameter counts
   • Only an alphabetical list of document filenames
   • Regular expressions and context-free grammars
3. Why is a simple finite-state machine not a complete description of every general nested context-free grammar decoder?
   • General nested structure can require stack-like or richer parser state
   • All nested grammars must be regular expressions by definition
   • A grammar can never describe nesting
   • Every parser must discard its current state after each token
4. Каква е основната техническа трудност при прилагането на ограничения на ниво токен?
   • Моделът моментално свършва паметта
   • Границите на токенизатора може да не съответстват на граматическите граници
   • Граматиките не могат да бъдат написани за JSON
   • Графичните процесори не могат да умножават матрици
5. What structure claim is justified for an output that successfully completes under a correctly implemented supported grammar constraint?
   • Every fact inside it is necessarily true
   • Every possible interruption leaves a complete accepted output
   • Every downstream application rule is automatically covered
   • It satisfies the constraints covered by that grammar
6. Which named research engine studies flexible and efficient structured generation for language models?
   • A photo-upscaling file format
   • An audio loudness unit
   • XGrammar
   • A required model-weight encryption algorithm
7. Generation stops after only an opening JSON object brace. What can be concluded about that text?
   • It must already be a complete JSON object
   • It can be a permitted prefix without being a complete valid object
   • The interruption proves no JSON grammar exists
   • Every such prefix is necessarily forbidden from the start
8. A stated distribution assigns A 0.2, B 0.3, and C 0.5. After C is masked and the remaining choices are renormalized, what is A's probability?
   • 0.4
   • 0.2
   • 0,5
   • 0.6