Quiz: Module 11 - Advanced Topics

Question 1

What is Mixture of Experts (MoE)?

1. Architecture with multiple expert networks and a router
2. Team of human experts
3. Ensemble of different models
4. Expert system AI

Question 2

In Mixtral, how many experts are active per token?

1. 2 out of 8
2. All 8
3. 1 out of 8
4. 4 out of 8

Question 3

What is the router's role in MoE?

1. Select which experts process each token
2. Route network traffic
3. Optimize training
4. Handle user requests

Question 4

What is GPT-4V?

1. GPT-4 with vision capabilities
2. GPT-4 version number
3. Video generation model
4. Voice model

Question 5

What vision encoder does LLaVA use?

1. CLIP
2. ResNet
3. ViT only
4. Custom encoder

Question 6

What is ReAct pattern?

1. Reasoning + Acting - alternating thought and action
2. React.js framework
3. Reaction time testing
4. Reactive programming

Question 7

What is expert collapse in MoE?

1. Router sending most tokens to few experts
2. System crash
3. Memory overflow
4. Training failure

Question 8

What is an LLM agent?

1. LLM that can plan, use tools, and take actions
2. Customer service bot
3. API wrapper
4. Training assistant

Question 9

What is tool use in LLMs?

1. LLM calling external functions/APIs
2. Using development tools
3. Hardware tools
4. Debugging tools

Question 10

What is Gemini's multimodal approach?

1. Native multimodal - trained on mixed modalities from start
2. Separate models combined
3. Text-only with image adapter
4. Vision-only model

Questions 11-30

(Covering: MoE training challenges, multimodal alignment, agent frameworks, tool safety, sparse attention, vision transformers, multi-agent systems, memory in agents, planning capabilities, grounding, action spaces, reward design for agents, failure modes, observability, scalable oversight)

Question 11

What is MoE's main efficiency benefit?

1. Higher capacity with same compute (sparse activation)
2. Simpler architecture
3. Less training data
4. Easier deployment

Question 12

What is load balancing in MoE?

1. Ensuring all experts get similar workload
2. Server load balancing
3. Battery management
4. Memory allocation

Question 13

What is image-text alignment?

1. Training visual and text representations in shared space
2. Aligning images on page
3. Text wrapping around images
4. Caption positioning

Question 14

What is LangChain?

1. Framework for building LLM applications with chains and agents
2. Translation tool
3. Language detection
4. Blockchain for languages

Question 15

What safety concern exists with agents?

1. Uncontrolled actions and tool misuse
2. Too slow responses
3. High costs only
4. Poor grammar

Question 16

What is vision transformer (ViT)?

1. Transformer architecture for images using patches
2. Image editing tool
3. Visual programming
4. Screen capture

Question 17

What is multi-agent collaboration?

1. Multiple LLM agents working together on tasks
2. Multiple users chatting
3. Distributed training
4. Parallel inference

Question 18

What is memory in agent context?

1. Storing past interactions and retrieved information
2. RAM usage
3. Model weights
4. Cache memory

Question 19

What is grounding in multimodal LLMs?

1. Connecting language to visual entities
2. Electrical grounding
3. Baseline establishment
4. Safety grounding

Question 20

What is action space in agents?

1. Set of possible actions agent can take
2. Physical space
3. Memory space
4. Disk space

Question 21

What is AutoGPT?

1. Autonomous agent that pursues goals with minimal supervision
2. Automatic training
3. Self-updating GPT
4. Car AI

Question 22

What is a hallucination in vision models?

1. Describing objects not present in image
2. Generating images
3. Visual effects
4. Image distortion

Question 23

What is the observation in ReAct?

1. Result returned from executing an action
2. User watching
3. Monitoring system
4. Scientific observation

Question 24

What is sparse attention?

1. Attending to subset of positions instead of all
2. Low attention span
3. Distributed attention
4. Weak attention

Question 25

What is sandboxed execution?

1. Running code in isolated environment for safety
2. Beach simulation
3. Testing environment
4. Development mode

Question 26

What is CrewAI?

1. Framework for multi-agent collaboration
2. Ship crew management
3. Team communication tool
4. HR software

Question 27

What is the projection layer in LLaVA?

1. Aligns vision encoder output to LLM input dimension
2. Display projection
3. Future prediction
4. Data projection

Question 28

What is planning in agents?

1. Breaking goals into steps and ordering actions
2. Schedule management
3. Resource allocation
4. Budget planning

Question 29

What is scalable oversight?

1. Maintaining human supervision as AI capabilities grow
2. Scaling monitoring systems
3. Large-scale supervision
4. Distributed management

Question 30

What is tool permission in agents?

1. Controlling which actions/tools agent can access
2. Software licensing
3. User permissions
4. File permissions

---

Passing Score: 24/30 (80%)

← Chapter 3Module 12 →