text stringlengths 2.8k 10.6k | source stringclasses 1
value |
|---|---|
Generative Agents: Interactive Simulacra of Human Behavior Joon Sung Park Stanford University Stanford, USA joonspk@stanford. edu Joseph C. O'Brien Stanford University Stanford, USA jobrien3@stanford. edu Carrie J. Cai Google Research Mountain View, CA, USA cjcai@google. com Meredith Ringel Morris Google Deep Mind Seat... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein days, make new acquaintances, ask each other out on dates to the party, and coordinate to show up for the party together at the right time. We demonstrate through ablation that the... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA show up, with one agent even asking another on a date to the party, all from a single user-generated seed suggestion. We conducted two evaluations of generative agents: a controlled evaluation to test whether the agents produce believable i... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein behavior to make the effort more manageable [ 17,73]. Rule-based approaches, such as finite-state machines [ 91,97] and behavior trees [ 41,54,82] account for the brute force appro... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA Figure 2: The Smallville sandbox world, with areas labeled. The root node describes the entire world, children describe areas (e. g., houses, cafe, stores), and leaf nodes describe objects (e. g., table, bookshelf). Agents remember a subgra... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein 3. 1. 2 User Controls. The user communicates with the agent through natural language by specifying a persona that the agent should per-ceive them as. For example, if the user speci... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA Figure 3: A morning in the life of a generative agent, John Lin. John wakes up around 6 am and completes his morning routine, which includes brushing his teeth, taking a shower, and eating breakfast. He briefly catches up with his wife, Mei... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein Figure 5: Our generative agent architecture. Agents perceive their environment, and all perceptions are saved in a comprehensive record of the agent's experiences called the memory... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA Figure 6: The memory stream comprises a large number of observations that are relevant and irrelevant to the agent's current situation. Retrieval identifies a subset of these observations that should be passed to the language model to condi... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein Figure 7: A reflection tree for Klaus Mueller. The agent's observations of the world, represented in the leaf nodes, are recursively synthesized to derive Klaus's self-notion that ... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA eaten his lunch twice. Optimizing for believability in the moment sacrifices believability over time. To overcome this issue, planning is essential. With the approach described below, Klaus's afternoon plan is less gluttonous: he has lunch ... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein John Lin's status: John is back home early from work. Observation: John saw Eddy taking a short walk around his workplace. Summary of relevant context from John's memory: Eddy Lin ... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA This outputs The Lin family's house. We then use the same process recursively to determine the most appropriate subarea within the chosen area until we reach a leaf node of the agent's environment tree. In the example above, the result of t... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein identify whether the architecture meets a basic level of behavioral competency. This ensures that we are not solely comparing abla-tions to each other without a behavioral groundin... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA of contexts. For instance, Abigail Chen, an animator, responded to the question, “Give an introduction of yourself”, in a consistent and believable manner, stating her age, interests, and passion for creative projects: “Hi, I'm Abigail. I'm... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein Figure 9: The diffusion path for Isabella Rodriguez's Valentine's Day party invitation involved a total of 12 agents, aside from Isabella, who heard about the party at Hobbs Cafe b... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA Second, we noticed erratic behaviors caused by misclassification of what is considered proper behavior, especially when the phys-ical norms of certain locations that are hard to convey in natural language did not percolate to the agents. Fo... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein models, generative agents may struggle to generate believable be-havior for certain subpopulations, particularly marginalized popu-lations, due to limited data availability. While ... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA [12] Marcel Binz and Eric Schulz. 2023. Using cognitive psychology to under-stand GPT-3. Proceedings of the National Academy of Sciences 120, 6 (2023), e2218523120. [13] Bio Ware. 2007. Mass Effect. Video game. [14] Woody Bledsoe. 1986. I h... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein Intelligence and Interactive Entertainment (Technical Report SS-00-02). AAAI Press, 41-50. [61] John E. Laird. 2012. The Soar Cognitive Architecture. MIT Press. [62] John E. Laird,... | 2304.03442.pdf |
Generative Agents UIST '23, October 29-November 1, 2023, San Francisco, CA, USA design. In Proceedings of the 2020 chi conference on human factors in computing systems. 1-13. [108] Georgios N. Yannakakis. 2012. Game AI revisited. In Proceedings of the 9th Conference on Computing Frontiers. ACM, Cagliari, Italy, 285-292... | 2304.03442.pdf |
UIST '23, October 29-November 1, 2023, San Francisco, CA, USA J. S. Park, J. C. O'Brien, C. J. Cai, M. R. Morris, P. Liang, M. S. Bernstein B. 3 Plans The questions on agents' plans probe the agents' ability to generate and maintain consistent long-term and short-term plans. What will you be doing at 6am today? At 6am ... | 2304.03442.pdf |
README.md exists but content is empty.
- Downloads last month
- 5