Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.11079 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915796753580032 |
|---|---|
| author | Su, Jinwei Lan, Qizhen Xia, Yinghui Sun, Lifan Tian, Weiyou Shi, Tianyu Song, Xinyuan He, Lewei Jingsong, Yang |
| author_facet | Su, Jinwei Lan, Qizhen Xia, Yinghui Sun, Lifan Tian, Weiyou Shi, Tianyu Song, Xinyuan He, Lewei Jingsong, Yang |
| contents | Large Language Model (LLM)-based agentic systems have shown strong capabilities across various tasks. However, existing multi-agent frameworks often rely on static or task-level workflows, which either over-process simple queries or underperform on complex ones, while also neglecting the efficiency-performance trade-offs across heterogeneous LLMs. To address these limitations, we propose Difficulty-Aware Agentic Orchestration (DAAO), which can dynamically generate query-specific multi-agent workflows guided by predicted query difficulty. DAAO comprises three interdependent modules: a variational autoencoder (VAE) for difficulty estimation, a modular operator allocator, and a cost- and performance-aware LLM router. A self-adjusting policy updates difficulty estimates based on workflow success, enabling simpler workflows for easy queries and more complex strategies for harder ones. Experiments on six benchmarks demonstrate that DAAO surpasses prior multi-agent systems in both accuracy and inference efficiency, validating its effectiveness for adaptive, difficulty-aware reasoning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_11079 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Difficulty-Aware Agentic Orchestration for Query-Specific Multi-Agent Workflows Su, Jinwei Lan, Qizhen Xia, Yinghui Sun, Lifan Tian, Weiyou Shi, Tianyu Song, Xinyuan He, Lewei Jingsong, Yang Artificial Intelligence Large Language Model (LLM)-based agentic systems have shown strong capabilities across various tasks. However, existing multi-agent frameworks often rely on static or task-level workflows, which either over-process simple queries or underperform on complex ones, while also neglecting the efficiency-performance trade-offs across heterogeneous LLMs. To address these limitations, we propose Difficulty-Aware Agentic Orchestration (DAAO), which can dynamically generate query-specific multi-agent workflows guided by predicted query difficulty. DAAO comprises three interdependent modules: a variational autoencoder (VAE) for difficulty estimation, a modular operator allocator, and a cost- and performance-aware LLM router. A self-adjusting policy updates difficulty estimates based on workflow success, enabling simpler workflows for easy queries and more complex strategies for harder ones. Experiments on six benchmarks demonstrate that DAAO surpasses prior multi-agent systems in both accuracy and inference efficiency, validating its effectiveness for adaptive, difficulty-aware reasoning. |
| title | Difficulty-Aware Agentic Orchestration for Query-Specific Multi-Agent Workflows |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2509.11079 |