Research map / Multi-agent systems and games / LLM agents
DAWN: Distributed LLM Multi-Agent Workflow Synthesis
Guancheng Wan, Mo Zhou, Ziyi Wang, Xiaoran Shang, Eric Hanchen Jiang, Guibin Zhang, Jinhe Bi, Yunpu Ma, Zaixi Zhang, Ke Liang, Wenke Huang
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 DAWN, a method for synthesizing multi-agent communication workflows (graphs) in a distributed setting where data is siloed across organizations due to privacy constraints.
Problem
Designing adaptive communication graphs for LLM multi-agent systems is hard when privacy constraints silo data across organizations, so workflows cannot be designed from centralized data.
Approach
Synthesizes multi-agent workflows in a distributed manner across data silos rather than with centralized access to all data; the abstract is truncated so specific mechanism details are not stated.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Parametric Resonance (SVD-based aggregation) (ablation) | DAWN · GSM8K (K=5) | accuracy: +0.58% | Improved |
| add Parametric Resonance (SVD-based aggregation) (ablation) | DAWN · MMLU (K=5) | accuracy: +1.27% | Improved |
| add Structural Gravity (FGW regularization) (ablation) | DAWN · GSM8K (K=5) | accuracy: +0.44% | Improved |
| add Structural Gravity (FGW regularization) (ablation) | DAWN · MMLU (K=5) | accuracy: +1.03% | Improved |
| combine both components (ablation) | DAWN · GSM8K / MMLU (K=5) | accuracy: +1.25% GSM8K, +3.13% MMLU | Improved |
| DAWN vs FedProx average | FedProx · 3 datasets | average accuracy: 87.05 -> 88.07 | Improved |
| DAWN vs AFlow (centralized agent baseline) | AFlow · 3 datasets | average accuracy: 87.77 -> 88.07 | Improved |
| DAWN vs FedProx on HumanEval K=5 | FedProx · HumanEval (K=5) | Pass@1: 91.47 -> 93.02 | Improved |
| increase agent count 5 to 15 (ablation) | DAWN · MMLU (K=3, K=5) | accuracy: approximately 4% gain | Improved |
Closest papers by meaning
- GraphFlow: A Graph-Based Workflow Management for Efficient LLM-Agent Serving (ICML 2026 · similarity 0.89)
- AgentNet: Decentralized Evolutionary Coordination for LLM-based Multi-Agent Systems (NeurIPS 2025 · similarity 0.89)
- Benchmarking Agentic Workflow Generation (ICLR 2025 · similarity 0.89)
- Graph-of-Agents: A Graph-based Framework for Multi-Agent LLM Collaboration (ICLR 2026 · similarity 0.89)
- Optimize Once, Execute Fast: Latency-Aware Multi-Agent Workflow Learning for Recurrent Queries (NeurIPS 2026 · similarity 0.89)
- Flow: Modularized Agentic Workflow Automation (ICLR 2025 · similarity 0.88)
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.
