Research map / Multi-agent systems and games / LLM agents
Many Minds, One Path: LLM-Augmented Consensus Decision for Distributed Control in Multi-Agent Collaborative Stable Scenarios
Zhuohao Yu, Zhe Liu, Tao Ren, Chenxue Wang, Junjie Wang, Qing Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes LLMASC, a framework that uses LLM-augmented consensus decisions to improve long-term stability of distributed multi-agent control in collaborative stable scenarios.
Problem
Long-term stability (coherent, safe system behavior over time) of distributed multi-agent systems in high-stakes settings such as power grids, traffic systems, and collaborative robotics is critical but underexplored.
Approach
Combines LLM reasoning with a consensus decision mechanism among distributed agents to enhance long-term stability in multi-agent collaborative control (details beyond the abstract not available).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Semantic Perception Encoder (ablation) | LLMASC · MultiWalker | Stability rate: 36% -> 78% | Improved |
| add Semantic Perception Encoder (ablation) | LLMASC · MultiWalker | Long-term success rate: 54% -> 80% | Improved |
| add Semantic Perception Encoder (ablation) | LLMASC · SUMO | Stability rate: 74% -> 93% | Improved |
| add Semantic Perception Encoder (ablation) | LLMASC · SUMO | Long-term success rate: 76% -> 98% | Improved |
| add Semantic Perception Encoder (ablation) | LLMASC · PowerGrid | Stability rate: 44% -> 75% | Improved |
| add Semantic Perception Encoder (ablation) | LLMASC · PowerGrid | Long-term success rate: 52% -> 78% | Improved |
| replace GPT-4o with small Qwen3-0.6B (ablation) | LLMASC · MultiWalker | Long-term success rate: 80% -> 74% | Mixed |
| replace GPT-4o with larger proprietary LLMs (Claude-3.5) (ablation) | LLMASC · MultiWalker | Long-term success rate: 80% -> 76% | No clear effect |
| use HAPPO instead of MAPPO as base RL (ablation) | LLMASC · MultiWalker | Long-term success rate: 80% -> 64% | Hurt |
| use HASAC instead of MAPPO as base RL (ablation) | LLMASC · MultiWalker | Long-term success rate: 80% -> 64% | Hurt |
| use HAPPO instead of MAPPO as base RL (ablation) | LLMASC · PowerGrid | Stability rate: 75% -> 71% | Mixed |
| increase agent count from 3 to 9 | LLMASC · MultiWalker | Stability rate: 78% -> 72% | Hurt |
Closest papers by meaning
- Dr. MAS: Stable Reinforcement Learning for Multi-Agent LLM Systems (NeurIPS 2026 · similarity 0.87)
- Iterative Critique-and-Routing Controller for Multi-Agent Systems with Heterogeneous LLMs (NeurIPS 2026 · similarity 0.87)
- ResMAS: Resilience Optimization in LLM-based Multi-agent Systems (AAAI 2026 · similarity 0.86)
- Adaptive Collaboration with Humans: Metacognitive Policy Optimization for Multi-Agent LLMs with Continual Learning (ICLR 2026 · similarity 0.86)
- $\texttt{Multi}^2$: Hierarchical Multi-Agent Decision-Making with LLM-Based Agents in Interactive Environments (ICML 2026 · similarity 0.86)
- SafeSieve: From Heuristics to Experience in Progressive Pruning for LLM-based Multi-Agent Communication (AAAI 2026 · 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.
