Research map / Multi-agent systems and games / Cooperative multi-agent RL
Neurosymbolic Transformers for Multi-Agent Communication
Jeevana Priya Inala, Yichen Yang, James Paulos, Yewen Pu, Osbert Bastani, Vijay Kumar, Martin Rinard, Armando Solar-Lezama
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a synthesized programmatic communication policy with transformer retraining instead of fixed distance-based communication | Neurosymbolic transformer · random-cross and random-grid | maximum communication degree | Improved |
| Use a synthesized programmatic communication policy with transformer retraining instead of fixed distance-based communication | Neurosymbolic transformer · random-cross and random-grid | loss | No clear effect |
| Use programmatic communication instead of thresholding transformer attention | Neurosymbolic transformer · random-cross and random-grid | maximum out-degree | Improved |
| Retrain the transformer after synthesizing the communication program (ablation) | Transformer + program · random-cross and random-grid | loss | Improved |
| Use a synthesized programmatic communication policy instead of fixed distance-based communication | Neurosymbolic transformer · unlabeled goals | maximum communication degree | Improved |
| Use the synthesized programmatic communication policy | Neurosymbolic transformer · unlabeled goals | loss | Hurt |
| Allow random-choice rules in the program search space (ablation) | Programmatic communication policy · random-cross and random-grid | maximum out-degree | Improved |
| Allow random-choice rules in the program search space (ablation) | Programmatic communication policy · unlabeled goals | loss | Improved |
| Use decision-tree communication policies instead of list-operating programs | Learned communication policy · random-cross, random-grid, and unlabeled goals | communication degree | Hurt |
| Use policy-gradient communication decisions as actions instead of teacher-guided program synthesis | Communication policy · multi-agent planning tasks | loss and communication degree | Mixed |
Closest papers by meaning
- Learning Multi-Agent Communication from Graph Modeling Perspective (ICLR 2024 · similarity 0.89)
- Bridging Training and Execution via Dynamic Directed Graph-Based Communication in Cooperative Multi-Agent Systems (AAAI 2025 · similarity 0.89)
- Expressive Multi-Agent Communication via Identity-Aware Learning (AAAI 2024 · similarity 0.88)
- Learning Efficient and Robust Multi-Agent Communication via Graph Information Bottleneck (AAAI 2024 · similarity 0.87)
- Towards Efficient Collaboration via Graph Modeling in Reinforcement Learning (AAAI 2025 · similarity 0.87)
- Multi-Agent Reinforcement Learning with Communication-Constrained Priors (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
