BusMA introduces a bus-based communication substrate for multi‑agent systems, enabling any agent to address others through a shared channel. The framework includes agent registration, message routing, and shared memory management, and defines four communication intents—discussion, challenge, guidance, and request for explanation—to facilitate fine‑grained interaction. Experiments with two leading LLMs on 13 tasks in visual reasoning, mathematical reasoning, and knowledge retrieval show that BusMA consistently outperforms existing hierarchical manager‑worker and router‑based message passing methods.
arXiv:2609.05774v1 Announce Type: new Abstract: Recent multi-agent LLM systems increasingly rely on graph-structured communication to coordinate specialized agents. We revisit multi-agent orchestrati...
arXiv:2607. 05477v1 Announce Type: cross Abstract: Improving the task performance of Large Language Models (LLMs) is essential, yet scaling these models faces significant challenges such as diminishing returns and high costs.
arXiv:2607. 09330v1 Announce Type: new Abstract: Embodied agent teams powered by heterogeneous large language models (LLMs) are being widely deployed in physical artificial intelligence such as smart factories, warehouses, and service robotics.
arXiv:2608.30672v1 Announce Type: new Abstract: Recent advances in large language models and multimodal models have pushed remote sensing (RS) processing from simple perception models to agentic syst...
arXiv:2607. 23678v1 Announce Type: new Abstract: Large language models (LLMs) enable autonomous agents for reasoning, planning, and tool use.
arXiv:2608. 05729v1 Announce Type: new Abstract: As capabilities rapidly increase, AI agents can move from running inside one app to acting across a user's devices over time.
MemoryArena is a new evaluation gym that benchmarks agent memory in interdependent multi‑session tasks. Unlike prior benchmarks that test memorization or single‑session action in isolation, MemoryArena requires agents to acquire memory while interacting with the environment and then use that memory to guide future decisions across a range of tasks such as web navigation, planning, information search, and formal reasoning. The benchmark reveals that agents excelling on existing long‑context memory tests perform poorly here, highlighting a gap in current memory evaluation methods.
The paper argues that the architecture of multi‑agent large language model (LLM) frameworks, rather than just the intelligence of the underlying models, largely determines system performance. It introduces a taxonomy of architectural dimensions—such as orchestration, memory, planning interfaces, specialization, and communication topology—and presents MAFBench, a unified evaluation suite. An empirical study across nine frameworks, keeping the LLM constant, reveals six design principles and shows that choices like orchestration and communication topology can dramatically affect latency, accuracy, and coordination success.
arXiv:2606. 05304v1 Announce Type: new Abstract: Multi-agent systems (MAS) built on large language models are typically organized around roles, pipelines, and turn schedules, while the content that agents pass to one another is often left as unconstrained natural language.
arXiv:2609.08977v3 Announce Type: replace-cross Abstract: In this work, we present Gander, a native multimodal duplex interaction model that builds on MiniCPM-o 4.5 and is further adapted for realtim...
arXiv:2609.35965v1 Announce Type: cross Abstract: Vision-and-Language Navigation (VLN) has largely focused on a single agent following a single instruction, yet many real-world applications require t...
The paper introduces AMA, a framework that uses multiple agents—Constructor, Retriever, Judge, and Refresher—to manage memory for large language model agents. AMA’s hierarchical memory design dynamically adjusts retrieval granularity to match task complexity, while the Judge and Refresher ensure relevance, consistency, and timely updates. Experiments on long-context benchmarks show AMA outperforms existing baselines and cuts token usage by about 80% compared to full-context approaches.