Skill-MAS: Evolving Meta-Skill for Automatic Multi-Agent Systems

arXiv:2606.18837 · cs.MA, cs.AI, cs.LG · Submitted 2026-06-17 · Read on arXiv

cs.MA, cs.AI, cs.LG

Submitted: 2026-06-17

Updated: 2026-09-29

Terminology

• Multi-Agent System — 2× in this paper · explained in 6 episodes
• Multi-Agent Systems — 2× in this paper
• Benchmark — 1× in this paper · explained in 2 episodes
• Large Language Model — 1× in this paper
• Large Language Model (LLM) — 1× in this paper · explained in 3 episodes
• Transferability — 1× in this paper · explained in 5 episodes

Sources

• EvoSkill: Automated Skill Discovery for Multi-Agent Systems
• BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent
• DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents
• From LLM Reasoning to Autonomous AI Agents: A Comprehensive Review
• VitaBench: Benchmarking LLM Agents with Versatile Interactive Tasks in Real-world Applications
• Automated Design of Agentic Systems
• AMA: Adaptive Memory via Multi-Agent Collaboration
• The Illusion of Multi-Agent Advantage
• A Survey of Frontiers in LLM Reasoning: Inference Scaling, Learning to Reason, and Agentic Systems
• MAS-Orchestra: Understanding and Improving Multi-Agent Reasoning Through Holistic Orchestration and Controlled Benchmarks
• MAS-ZERO: Designing Multi-Agent Systems with Zero Supervision
• When Single-Agent with Skills Replace Multi-Agent Systems and When They Fail
• ARISE: Agent Reasoning with Intrinsic Skill Evolution in Hierarchical Reinforcement Learning
• Interactive Learning for LLM Reasoning
• Unified-MAS: Universally Generating Domain-Specific Nodes for Empowering Automatic Multi-Agent Systems
• SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization
• Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills
• SkillMAS: Skill Co-Evolution with LLM-based Multi-Agent System
• Humanity's Last Exam
• PostTrainBench: Can LLM Agents Automate LLM Post-Training?

Related papers

• Highway Congestion Reduction through Reinforcement Learning Based Eulerian Headway Control
• ORACLE-SWE: Quantifying the Contribution of Critical Information Signals on SWE Agents
• You Only Align Once: Propagating Cooperative Behaviors in Multi-Agent Systems through Seed Agents
• Deny Without Disabling: Authorization-Paired Evaluation and Control for Multi-Agent Systems
• MA-SAPO: Multi-Agent Reasoning for Score-Aware Prompt Optimization
• PeroMAS: A Multi-agent System of Perovskite Material Discovery