All Topics
Topic135 questions
Accenture Generative AIQuestions & Answers
Practice 135 verified Accenture Generative AI questions with detailed answers and explanations. Tap any question below to study the full solution — perfect for last-minute Accenture primer and dumps prep.
Generative AI question list
Generative AI primarily aims to:Which of these is NOT typically produced by generative models?Learning a data distribution p(x) allows a model to:Which statement best contrasts discriminative and generative models?Which is a common application of generative AI?Generative AI that helps artists by suggesting concepts is an example of:A model that learns to produce plausible human faces has learned approximations of:Which capability is NOT typical of generative models?Which of the following is a risk specifically mentioned for generative AI?Text generation, image generation and music generation are examples of:Why is learning a distribution more powerful than memorizing examples?Which of these is a direct benefit of synthetic data?A generative model that outputs new molecules would be used in:Which term best describes creating content that resembles training data but is not identical?Generative AI differs from classification because it focuses on:Gaussian Mixture Models (GMMs) are examples of:Hidden Markov Models (HMMs) are especially useful for:Which breakthrough enabled deep generative models to scale in the 2010s?The VAE paper was published by:GANs introduced the idea of:“Attention Is All You Need” introduced:Which year is commonly associated with the original GAN paper?Transformers replaced recurrence with:Which early model is probabilistic and explicitly models density?VAEs are celebrated for:Which model family is known as “implicit density”?The rise of LLMs was enabled by:Which contribution is attributed to Goodfellow et al.?CycleGAN is notable because it can:Which development made sampling from complex distributions more practical?Machine Learning systems typically start with:A perceptron computes:Which activation is most used to mitigate vanishing gradients?Backpropagation uses which calculus tool to compute gradients?Gradient descent updates weights to:Deep networks learn hierarchical features—early layers learn:Overfitting happens when the model:Which is NOT an optimizer for neural networks?Dropout is used to:Cross-entropy loss is most often used for:A bias term in a neuron is analogous to:Batch normalization primarily helps by:Which layer type is most common in image models?Transfer learning helps when:An epoch means:Explicit density models provide:Normalizing Flows are an example of:Which model family does a VAE belong to?Implicit models are characterized by:Which is a tractable explicit model?Which approach approximates likelihoods using ELBO?Sampling from an implicit model requires:Which model gives exact likelihoods (when tractable)?Which is an advantage of explicit density models?An example of implicit modeling is:Which family is well-suited to likelihood-based anomaly detection?ELBO stands for:Which is a limitation of implicit models?Tractable models are useful because they allow:VAEs, GANs and Flows are examples of:A standard autoencoder differs from a VAE because a VAE:The reparameterization trick allows:VAE loss includes reconstruction loss plus:Sampling z = μ + σ ⊙ ε moves randomness to:A common prior used in VAEs is:VAEs typically produce images that are:KL term in VAE encourages:Advantages of VAEs include:Which is a limitation of VAEs?In VAEs, the decoder maps from:ELBO maximization is equivalent to:Choosing a too-large KL weight will typically:VAEs are useful for:A well-structured latent space allows:Which is true about VAE encoder output?GANs train by:Mode collapse means the generator:If discriminator becomes too strong early, the generator may suffer from:DCGAN stands for a GAN variant optimized for:StyleGAN introduced:CycleGAN is primarily used for:The generator maps noise z to:Adversarial loss tries to make discriminator output for generated samples:A typical fix for mode collapse is:GANs are categorized as:Which is a common component of GAN training to stabilize it?Which GAN variant gives control over style at multiple scales?Discriminator\'s role is to:GAN training objective is best described as:A challenge when training GANs is:RNNs maintain memory via:Vanishing gradient makes it hard to learn:LSTM introduces which mechanism to control information?GRU differs from LSTM by:Sequence generation can be performed by training models to predict:Teacher forcing is a training technique where:Which is a limitation of RNNs compared to Transformers?RNN backpropagation through time requires:Applications of sequence models include:Beam search is used in generation to:Scheduled sampling mixes:An RNN cell output depends on:Which cell is computationally lighter?Sequence-to-sequence (seq2seq) models typically have:Teacher forcing can lead to:Self-attention allows tokens to:Positional encoding provides:Multi-head attention helps by:Transformers are more parallelizable than RNNs because:Decoder-only models like GPT are trained to:BERT is primarily used for:Transformer encoder blocks include:Masked self-attention prevents a token from attending to:Scaling transformers (more params + data) led to:A positional encoding can be:Which model is decoder-only?Attention scores are computed from queries, keys and values using:Transformer attention is typically multi-head to:Encoder-decoder transformers are commonly used for:Which is an advantage of Transformers over RNNs?Generative AI in healthcare can help by:In drug discovery generative models can:A major ethical risk of generative AI is:Which practice helps reduce model bias?Copyright concerns arise because models may:Responsible deployment includes:Which industry widely uses generative AI for creative media?Data augmentation via generative models mainly helps to:Regulation and policy are needed because:A practical mitigation for deepfakes is:Multi-modal generative models combine:Job displacement risk suggests:Which direction is important for future generative AI?Intellectual property questions involve:When deploying a generative model for production, you should:Practice more Accenture topics
PrimerPrep has 1400+ primer questions, 2026 mocks and coding hands-on — all free.