Generative Agents and Simulated Societies Quiz

Answers and explanations appear as you take the quiz.

1. What famous behavior emerged on its own in Stanford's Smallville simulation?
   • An agent planned a Valentine's Day party and others spread the word and attended
   • Agents started a war over territory
   • Agents refused to leave their houses
   • All agents simultaneously crashed the server
2. What is the 'memory stream' in a generative agent?
   • A live video feed of the simulation
   • A timestamped log of everything the agent has observed
   • The GPU memory used to run the model
   • A list of pre-written dialogue lines
3. When an agent decides how to act, how are relevant memories chosen?
   • Randomly from the full log
   • Only the most recent memory is ever used
   • By a score combining recency, importance, and relevance
   • A human picks them manually
4. What does the 'reflection' step do in the agent architecture?
   • Renders the agent's face in 3D
   • Deletes old memories to save space
   • Pauses the simulation for review
   • Summarizes raw memories into higher-level insights
5. Why is planning important for believable agents?
   • It lets agents act consistently over long time spans instead of forgetting
   • It makes the graphics render faster
   • It encrypts the agent's data
   • It connects the agent to the internet
6. Which is a plausible real-world use of simulated agent societies?
   • Replacing all human elections
   • Pre-testing how a rumor or policy might spread before real deployment
   • Permanently storing human consciousness
   • Mining cryptocurrency more efficiently
7. How many generative agents were placed in the Smallville sandbox in the 2023 study?
   • 5
   • 100
   • 25
   • 1,000
8. Which model powered the Smallville agents described in the paper?
   • GPT-2
   • BERT
   • A hidden Markov model
   • GPT-3.5