Multi-Agent Coordination via Multi-Level Communication
Authors: Gang Ding, Zeyuan Liu, Zhirui Fang, Kefan Su, Liwen Zhu, Zongqing Lu
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, we show that Seq Comm outperforms existing methods in various cooperative multi-agent tasks. |
| Researcher Affiliation | Collaboration | 1Tsinghua Shenzhen International Graduate School, Tsinghua University, 2Peking University, 3Tencent AI Lab |
| Pseudocode | Yes | For more details of the algorithms, please refer to the Appendix D for the pseudo-code. |
| Open Source Code | No | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [No] Justification: We plan to release data and code ASAP/upon acceptance. |
| Open Datasets | Yes | Empirically, we evaluate Seq Comm on Star Craft multi-agent challenge v2 (SMACv2) [Samvelyan et al., 2019]. |
| Dataset Splits | No | The paper evaluates its method on SMACv2 maps but does not explicitly describe specific training, validation, and test dataset splits with percentages, sample counts, or references to predefined splits for their experimental setup. |
| Hardware Specification | Yes | We trained our model on one Ge Force GTX 1050 Ti and Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz. |
| Software Dependencies | No | The paper mentions that models are implemented based on MAPPO and that baselines are fine-tuned, but it does not specify exact version numbers for programming languages (e.g., Python), libraries (e.g., PyTorch, TensorFlow), or other software components used in the experiments. |
| Experiment Setup | Yes | For Protoss, the learning rate is 1e 5, while for Terran and Zerg, the learning rate is 2.5e 5. H and F for calculating intention value is set to 20 and 2. For Tar MAC, the learning rate is tuned as 5e 5. Tar MAC adopts MAPPO as the backbone and two-round communication mechanism. |