Generative Agents: Interactive Simulacra of Human Behavior

Generative Agents makes long-term memory part of an agent's control loop. Follow the shared memory stream, three-factor retrieval score, reflection tree, recursive plan, believability ablation, and the failures that separate stored history from usable context.

Reading focus: How recency, importance, and embedding relevance select a bounded prompt from a growing natural-language memory stream. Why reflections and plans are written back as memories, letting yesterday's conclusions shape today's reactions without changing model weights. What the five-condition believability study establishes, and why one two-day town, retrieval failures, paper-code drift, and a subjective metric bound the claim.

UIST 2023. Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 12 min read, medium difficulty.