LLM Midterm Quiz 2

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/11

encourage image

There's no tags or description

Looks like no tags are added yet.

Last updated 4:37 AM on 10/9/26
Name
Mastery
Learn
Test
Matching
Spaced
Call with Kai
Chat

No analytics yet

Send a link to your students to track their progress

12 Terms

1
New cards

Along with the special tokens [UNK], [SEP], [PAD], [CLS], [MASK], the cased version of BERT has another special token [CASE]. (T/F)

False

2
New cards

[PAD]

ensure all texts have the same length (longest text in the batch)

3
New cards

[BOS]

start of text

4
New cards

[EOS}

end of text

5
New cards

Rotary positional encoding captures absolute and relative token position information. (T/F)

True

6
New cards

BPE tokenizer can handle any unknown word by representing it as a sequence of ______ or ______ tokens


subword, characters

7
New cards

BERT uses which tokenization method?

WordPiece

8
New cards

Using a batch size of one during data loadinghas two advantages 1) uses less memory and 2) smoother update of the model. (T/F)

False

9
New cards

Which of the following is an easier tokenizer but difficult for modeling?

Byte Tokens

10
New cards

Which one of the following models uses the special token <work> - for chain of thought reasoning?

Galactica

11
New cards

Having context length (max_length) and stride equal to each other while defining a dataloader avoids overfitting. (T/F)

True

12
New cards

Which one of the following tokens was used to prepend independent text sources?

<endoftext>