dorsal/arxiv
View SchemaAdaptive Orchestration: Scalable Self-Evolving Multi-Agent Systems
| Authors | Sathish Sampath, Anuradha Baskaran |
|---|---|
| Categories | |
| ArXiv ID | 2601.09742vv1 |
| URL | https://arxiv.org/abs/2601.09742 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
As Large Language Models (LLMs) are increasingly deployed as autonomous agents, they face a critical scalability bottleneck known as the "Generalization-Specialization Dilemma." Monolithic agents equipped with extensive toolkits suffer from context pollution and attention decay, leading to hallucinations. Conversely, static multi-agent swarms introduce significant latency and resource overhead. This paper introduces a Self-Evolving Concierge System, a novel architecture utilizing a Dynamic Mixture of Experts (DMoE) approach. Unlike recent self-improving agents that rewrite their own codebase, our system preserves stability by dynamically restructuring its runtime environment: "hiring" specialized sub-agents based on real-time conversation analysis. We introduce an asynchronous "Meta-Cognition Engine" that detects capability gaps, a Least Recently Used (LRU) eviction policy for resource constraints, and a novel "Surgical History Pruning" mechanism to mitigate refusal bias. Experimental results demonstrate that this architecture maintains high task success rates while minimizing token consumption compared to static agent swarms.
{
"annotation_id": "1be323dc-5467-4f7b-a394-6b4962b9f92d",
"date_created": "2026-02-17T05:53:24.052000Z",
"date_modified": "2026-02-17T05:53:24.052000Z",
"file_hash": "05e127eae20011c5d30feb9f4fd2c6111af66137d3229ce985027b3b95fd41fd",
"private": false,
"record": {
"abstract": "As Large Language Models (LLMs) are increasingly deployed as autonomous agents, they face a critical scalability bottleneck known as the \"Generalization-Specialization Dilemma.\" Monolithic agents equipped with extensive toolkits suffer from context pollution and attention decay, leading to hallucinations. Conversely, static multi-agent swarms introduce significant latency and resource overhead. This paper introduces a Self-Evolving Concierge System, a novel architecture utilizing a Dynamic Mixture of Experts (DMoE) approach. Unlike recent self-improving agents that rewrite their own codebase, our system preserves stability by dynamically restructuring its runtime environment: \"hiring\" specialized sub-agents based on real-time conversation analysis. We introduce an asynchronous \"Meta-Cognition Engine\" that detects capability gaps, a Least Recently Used (LRU) eviction policy for resource constraints, and a novel \"Surgical History Pruning\" mechanism to mitigate refusal bias. Experimental results demonstrate that this architecture maintains high task success rates while minimizing token consumption compared to static agent swarms.",
"arxiv_id": "2601.09742",
"authors": [
"Sathish Sampath",
"Anuradha Baskaran"
],
"categories": [
"cs.MA"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "Adaptive Orchestration: Scalable Self-Evolving Multi-Agent Systems",
"url": "https://arxiv.org/abs/2601.09742",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "f047ddb9-f97a-45ba-8409-8780af630bc1",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}