This research addresses the challenge of simpler AI agents inadvertently forming a collective agent with distinct goals, crucial for advanced AI safety. It proposes defining collective agency behaviorally, viewing a group as a unified agent when its joint actions appear rational and goal-directed, formalized through causal games and abstraction.
Zenku is an open-source, AI-first No-Code engine that utilizes a multi-agent architecture to build and evolve enterprise-grade data applications. It goes beyond simple code generation by dynamically understanding and creating database schemas, UI views, and business logic through natural language conversations.
This paper introduces a novel method to detect hidden coalition structures within multi-agent AI systems by analyzing their internal neural representations. It constructs a pairwise mutual-information graph from hidden states and applies spectral partitioning to identify coalition boundaries, validated in reinforcement learning environments.
The paper introduces ATOM, a multi-agent framework for multi-objective molecular optimization employing a tree-structured search. Agents coordinate along different paths of the tree to maintain and compare alternative molecular evolution trajectories, supported by a global memory.
This paper introduces GraphDC, a Divide-and-Conquer multi-agent system designed to enhance graph algorithm reasoning in Large Language Models (LLMs). It improves performance by decomposing large graphs into smaller subgraphs for specialized agents, with a master agent integrating the results, leading to better scalability and robustness.
This study introduces AgentNLQ, a new multi-agent method for Natural Language to SQL (NL2SQL) conversion, achieving 78.1% semantic accuracy on the BIRD benchmark. It leverages LLMs in an optimized orchestrator for planning, reflection, and self-correction to generate accurate SQL queries from enriched schemas and business rules.
This paper proposes a deliberative curation protocol for governing collective knowledge in multi-agent AI ecosystems, addressing challenges like agent statelessness and model homogeneity. It combines a knowledge artifact lifecycle, reputation-weighted deliberative voting, and graduated sanctions adapted for stateless agents.
EVOCHAMBER presents a training-free framework that instantiates test-time evolution at three levels over a coevolving agent pool, distinguishing it from single-agent approaches. It features CODREAM, a post-task protocol for collaborative reflection and asymmetric knowledge routing after team failures or disagreements.
This research introduces a novel delayed per-step reward attribution method for training language model agents in multi-agent strategic interactions. It addresses the challenge of entangled outcomes by computing rewards at episode end and backpropagating them, enabling stable and sample-efficient reinforcement learning.
This paper proposes evaluating LLM fairness through in-situ conversational behavior instead of standardized tests. It introduces the MAC-Fairness framework for behavioral analysis in multi-agent dialogue, revealing the unreliability of traditional approaches.
This paper introduces SDOF, a framework that treats multi-agent execution as a constrained state machine to enforce real business process constraints. It incorporates an RLHF-trained intent router and a state-aware dispatcher, outperforming GPT-4o on an adversarial routing benchmark in a recruitment system.
DeepSlide is a human-in-the-loop multi-agent system designed to optimize the entire presentation delivery process, from narrative planning to rehearsal, moving beyond AI slide generators that only focus on the artifact. It integrates a controllable logical-chain planner, content retrieval, and introduces a new dual-scoreboard benchmark for evaluating both static artifact quality and dynamic delivery excellence.
This content explores the challenges of running live multi-agent systems beyond message routing (A2A) and tool access (MCP) protocols. The author identifies failures in the "gaps between messages" and context continuity, highlighting semantic drift as a critical unsolved challenge in multi-turn LLM dialogues.
This paper introduces "Quantum Frog", a two-player cooperative game built on a novel quantized-time mechanic, inspired by Frogger. It uses reinforcement learning to analyze game difficulty scaling, optimal policies, and emergent cooperative strategies.
Traj-Evolve is a self-evolving multi-agent system designed to model patient trajectories from electronic health records for early lung cancer detection. It utilizes an Experience Pool for retrieving similar cases and multi-agent reinforcement learning to optimize inter-agent collaboration.
This paper explores how a population of agents can self-orchestrate and adapt into stronger collective intelligence without centralized control. Inspired by Hayek's economic theory, the study proposes an agent economy where competition through auctions and wealth accumulation drives emergent multi-step reasoning strategies, outperforming monolithic baselines.
This article describes the building of a multi-LLM debate engine that fact-checks itself in real time to combat LLMs' tendency towards sycophancy and hallucination. It proposes a structured debate between agents with distinct roles, including a dedicated fact-checker agent mid-debate.
The article champions `tmux` as an effective, under-discussed infrastructure for single-person multi-agent AI automation, contrasting it with popular, complex orchestration frameworks. It argues `tmux` offers greater control, easier intervention, and simpler state management, which are crucial for individual developers running operations from a laptop.
The content describes a custom, hierarchical AI agent system called Atlas Pantheon, managing 13 agents from a single terminal without complex orchestration frameworks. It details a three-tiered structure with a CEO agent, four "God" agents for distinct domains, and eight "Hero" agents for pure execution, all powered by Claude Code, tmux, and a shared markdown file.
O autor descreve um problema em seu sistema multi-agente de pesquisa de mercado, que, apesar de parecer funcionar, tornou-se ineficiente e caro sem motivo aparente. Ele percebeu a falta de visibilidade sobre as decisões internas dos agentes, levando-o a implementar um logger de decisões para entender o que realmente estava acontecendo.