ProAgent: Building Proactive Cooperative Agents with Large Language Models
Authors: Ceyao Zhang, Kaijie Yang, Siyi Hu, Zihao Wang, Guanghe Li, Yihang Sun, Cheng Zhang, Zhaowei Zhang, Anji Liu, Song-Chun Zhu, Xiaojun Chang, Junge Zhang, Feng Yin, Yitao Liang, Yaodong Yang
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental evaluations conducted within the Overcooked-AI environment unveil the remarkable performance superiority of Pro Agent, outperforming five methods based on self-play and population-based training when cooperating with AI agents. |
| Researcher Affiliation | Academia | Ceyao Zhang1,2* , Kaijie Yang3*, Siyi Hu4*, Zihao Wang2,5, Guanghe Li2, Yihang Sun2, Cheng Zhang2, Zhaowei Zhang2,5, Anji Liu2, Song-Chun Zhu5, Xiaojun Chang4, Junge Zhang3, Feng Yin1, Yitao Liang2, Yaodong Yang2 1SSE, The Chinese University of Hong Kong, Shenzhen 2Institute for Artificial Intelligence, Peking University 3Institute of Automation, Chinese Academy of Sciences 4Re LER, AAII, University of Technology Sydney 5National Key Laboratory of General Artificial Intelligence, BIGAI |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states 'For more information about our project, please visit https://pku-proagent.github.io.', which is a project information page, not an explicit statement that the source code for the methodology is provided, nor a direct link to a code repository. |
| Open Datasets | Yes | To assess the adaptive cooperative capabilities of Pro Agent, we conducted performance evaluations using the well-established multi-agent coordination testing suite, Overcooked-AI (Carroll et al. 2019). |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training or validation data. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details, such as library names with version numbers, needed to replicate the experiment. |
| Experiment Setup | Yes | In the main experiments, we use L2 level prompts and recent-1 strategy. [...] we conducted an ablation study within the context of the Cramped Room layout. The experiment considered three distinct conditions and their respective scores were: 1) 204 for L3 level prompts (with both analysis and intention), 2) 184 for L2 level prompts (with analysis but no intention), and 3) 100 for L1 level prompts (making a skill plan directly, neither analysis nor intention). |