Official curriculum roadmap, subject/topic distribution, negative marking rules, pacing guidelines, and solved sample questions.
🎯 Mapped Subjects & Topic Question Distribution
Total Question Pool100%
82 MCQs
Combined Active Syllabus
Transformers, LLMs & Natural Language Processing
82 MCQs
Topic Pool
📊 Question Pool Structure
82 MCQs across fundamental, intermediate, and advanced concept tiers.
⚡ Recommended Pacing
45 to 60 seconds per MCQ. Flag complex problems and preserve 10 minutes for final revision.
⚖️ Scoring & Negative Marking
+1 mark per correct answer. In competitive tests with negative marking, -0.25 applies for incorrect guesses.
💡 Strategic Preparation & Exam Hall Guidelines
To maximize your score on Transformers, LLMs & Natural Language Processing, candidates are advised to follow a structured three-pass approach. In the First Pass, solve all direct recall and formula-based questions within 30 seconds each to secure foundational marks. In the Second Pass, tackle multi-step analytical and quantitative reasoning problems. In the Third Pass, review marked questions and verify calculations.
Practice with the interactive player below to evaluate your speed and accuracy under real exam pressure. Every question features full mathematical formulas, step-by-step worked solutions, and conceptual explanations vetted by Apex Rankers Academy subject matter specialists.
Ready to test your knowledge?Launch interactive 1-by-1 practice with instant feedback, bookmarking, and step-by-step rationales.
Review the solved problems below to understand question phrasing, answer choices, and step-by-step solution logic prior to starting the full interactive practice drill:
Sample Question 1
Transformers, LLMs & Natural Language ProcessingHard • Artificial Intelligence
What is the core mathematical formula for Scaled Dot-Product Attention in the original Transformer ('Attention Is All You Need', Vaswani et al. 2017)?
Dividing by sqrt(d_k) prevents dot products from growing excessively large for large dimensions, which would otherwise push the softmax into regions with vanishing gradients.
Sample Question 2
Transformers, LLMs & Natural Language ProcessingHard • Artificial Intelligence
Why is the scaling factor 1 / sqrt(d_k) crucial in Scaled Dot-Product Attention?
AFor large key dimensions d_k, the dot products grow large in magnitude, pushing the softmax function into regions with extremely small gradients
BIt reduces matrix memory size on GPU
CIt converts complex numbers to real floats
DIt bounds output values strictly between -1 and 1
✓ Correct Answer:A - For large key dimensions d_k, the dot products grow large in magnitude, pushing the softmax function into regions with extremely small gradients
📖 Step-by-Step Solution & Conceptual Rationale:
Under independent zero-mean unit-variance components, the dot product has mean 0 and variance d_k; dividing by sqrt(d_k) restores unit variance.
Sample Question 3
Transformers, LLMs & Natural Language ProcessingMedium • Artificial Intelligence
In Multi-Head Attention, why are multiple projection heads used rather than a single large attention head?
AMultiple heads allow the model to jointly attend to information from different representation subspaces at different positions simultaneously
BMultiple heads allow the model to train on multiple GPUs without code changes
CMultiple heads reduce the total number of matrix multiplications
DMultiple heads eliminate the need for feed-forward layers
✓ Correct Answer:A - Multiple heads allow the model to jointly attend to information from different representation subspaces at different positions simultaneously
📖 Step-by-Step Solution & Conceptual Rationale:
Multi-Head Attention: Concat(head_1, ..., head_h) * W_O, where each head can capture distinct syntactic and semantic relationships.
Sample Question 4
Transformers, LLMs & Natural Language ProcessingMedium • Artificial Intelligence
Why are 'Positional Encodings' strictly necessary in Transformer architectures?
ABecause the self-attention mechanism is inherently permutation-equivariant (it has no built-in notion of token order or sequence position)
BTo compress long sentences into short embeddings
CTo translate text between different natural languages
DTo prevent integer overflow during softmax calculation
✓ Correct Answer:A - Because the self-attention mechanism is inherently permutation-equivariant (it has no built-in notion of token order or sequence position)
📖 Step-by-Step Solution & Conceptual Rationale:
Unlike RNNs which process tokens step-by-step, Transformers process all tokens in parallel; without positional encodings, word order is lost.
Sample Question 5
Transformers, LLMs & Natural Language ProcessingHard • Artificial Intelligence
What is 'Byte-Pair Encoding' (BPE) tokenization widely used in GPT models?
AA subword tokenization algorithm that iteratively merges the most frequent pair of adjacent bytes or characters in a text corpus into a new vocabulary token
BAn encryption algorithm that secures text against cyberattacks
CA method that converts text into raw ASCII binary bits
DA tokenization method that splits text strictly on whitespaces
✓ Correct Answer:A - A subword tokenization algorithm that iteratively merges the most frequent pair of adjacent bytes or characters in a text corpus into a new vocabulary token
📖 Step-by-Step Solution & Conceptual Rationale:
BPE balances word-level semantics with character-level coverage, eliminating Out-Of-Vocabulary (OOV) tokens by decomposing rare words into subword pieces.
Sample Question 6
Transformers, LLMs & Natural Language ProcessingMedium • Artificial Intelligence
What is 'WordPiece' tokenization (used in BERT)?
AA subword tokenization algorithm that selects symbol pairs to merge based on maximizing the likelihood of the training language model rather than pure frequency count
BA tool for checking spelling mistakes in Microsoft Word
CA method that limits vocabularies to 1,000 words only
DA syntax parser that extracts noun phrases
✓ Correct Answer:A - A subword tokenization algorithm that selects symbol pairs to merge based on maximizing the likelihood of the training language model rather than pure frequency count
📖 Step-by-Step Solution & Conceptual Rationale:
WordPiece uses the '##' prefix for subword continuations (e.g. 'playing' -> 'play', '##ing') and optimizes corpus likelihood.
Sample Question 7
Transformers, LLMs & Natural Language ProcessingHard • Artificial Intelligence
What is 'SentencePiece' tokenization (used in T5, LLaMA)?
AA language-independent subword tokenizer that treats raw input text as a stream of Unicode characters including whitespace (represented as '_'), without requiring language-specific pre-tokenizers
BA model that generates full sentences from single words
CA grammar checking rule engine
DA tool that counts the number of sentences in a document
✓ Correct Answer:A - A language-independent subword tokenizer that treats raw input text as a stream of Unicode characters including whitespace (represented as '_'), without requiring language-specific pre-tokenizers
📖 Step-by-Step Solution & Conceptual Rationale:
SentencePiece enables seamless multi-lingual tokenization without language-specific regex or whitespace assumptions.
Sample Question 8
Transformers, LLMs & Natural Language ProcessingMedium • Artificial Intelligence
What is the fundamental architectural difference between BERT and GPT?
ABERT is an Encoder-only model trained with bidirectional masked language modeling; GPT is a Decoder-only autoregressive model trained with causal left-to-right next-token prediction
BBERT is for images; GPT is for audio
CBERT has no attention layers; GPT uses only attention layers
DBERT runs only on CPUs; GPT runs on TPUs
✓ Correct Answer:A - BERT is an Encoder-only model trained with bidirectional masked language modeling; GPT is a Decoder-only autoregressive model trained with causal left-to-right next-token prediction
📖 Step-by-Step Solution & Conceptual Rationale:
BERT excels at representation and comprehension (classification, NER, QA); GPT excels at generative text synthesis.
Sample Question 9
Transformers, LLMs & Natural Language ProcessingMedium • Artificial Intelligence
What is 'Masked Language Modeling' (MLM) pre-training objective introduced by BERT?
ARandomly masking 15% of input tokens with a [MASK] token and training the bidirectional encoder to predict the original identity of the masked words from context
BHiding user personal data for privacy
CMasking convolutional feature maps with zero padding
DFiltering out offensive words from internet text
✓ Correct Answer:A - Randomly masking 15% of input tokens with a [MASK] token and training the bidirectional encoder to predict the original identity of the masked words from context
📖 Step-by-Step Solution & Conceptual Rationale:
MLM enables deep bidirectional representation learning by forcing the model to fuse left and right context representations simultaneously.
Sample Question 10
Transformers, LLMs & Natural Language ProcessingHard • Artificial Intelligence
What is 'Causal Masking' (Autoregressive Masking / Look-Ahead Mask) in Transformer Decoders (e.g. GPT)?
AAn upper-triangular attention mask that sets future token attention scores to -infinity before softmax, ensuring position i can only attend to positions <= i
BA mask that prevents models from generating offensive text
CA mask that hides images from text decoders
DA technique that deletes punctuation tokens
✓ Correct Answer:A - An upper-triangular attention mask that sets future token attention scores to -infinity before softmax, ensuring position i can only attend to positions <= i
📖 Step-by-Step Solution & Conceptual Rationale:
Causal masking enforces the autoregressive property during parallel training: token t cannot cheat by looking at ground-truth future token t+1.
Practice All 82 Questions InteractivelyTest your knowledge in real-time with continuous progress saving, instant scoring, and performance analytics.
Loading questions from the question bank...
No questions found matching your selected filters.