Fri 13 Nov
LiveGet it free
ZOTPAPER

News without the noise

← All concepts

multi-agent architecture

Think of a multi-agent architecture as a team of specialized workers, each with their own role, who collaborate to solve a bigger problem. Instead of one giant program trying to do everything, you have smaller, focused programs, agents, that handle subtasks like planning, gathering information, or checking results. They communicate and share a common workspace, just like people on a project team, allowing them to tackle complex jobs more flexibly and robustly than a single, monolithic system.

Across these papers, multi-agent systems consistently outperform single-agent or simpler baselines, whether generating 3D scenes, videos, or clinical trial plans. The common thread is decomposing hard problems into stages managed by different agents, often with memory loops for revision and coordination without a single bottleneck. This approach improves success rates, consistency, and realism by letting specialists handle what they do best and collaborate effectively.

Papers
110
Released code
10
First seen
Apr 2026
Latest
Nov 2026

91 papers in the last two months, against 11 in the two before.

Start here

Three papers to read first, in the order they came out
  1. 1Adaptive interaction routing improves multimodal AI debate accuracy while cutting token use

    AIM boosts multimodal debate accuracy by adaptively choosing text, graph, or region-based interaction per query.

    Nov 2026
  2. 2AI agent teams consume more resources without a consistent advantage

    There is no universally best agent architecture; the right choice depends on the SOP and implementation quality.

    Nov 2026
  3. 3AI agents remain vulnerable to prompt injection when attackers use collaborating swarms

    Prompt injection attacks remain effective against frontier AI agents when attackers use swarms of collaborating LLMs.

    Nov 2026

How fast it is moving

JunJulAugSeptOctNov so far

Often studied with

large language modelA neural network trained on massive text corpora to generate coherent text and perform diverse language tasks.22 togetheragent memoryA repository where an agent stores past interactions and outcomes to inform future decisions.21 togethermulti-agent coordinationThe collaborative operation of multiple AI agents to solve tasks more effectively or with greater capability.16 together

Who is working on it

University of Illinois Urbana-Champaign 5Carnegie Mellon University 5Amazon 4Google DeepMind 4Google 4National University of Singapore 4

The papers

Most central to this idea first, not most recent.

Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.