Difficulty-Aware Agentic Orchestration for Query-Specific Multi-Agent Workflows
This addresses the efficiency-performance trade-off in multi-agent LLM systems for AI applications, though it is incremental as it builds on existing agentic frameworks.
The paper tackled the problem of static multi-agent workflows that inefficiently handle queries of varying complexity by proposing Difficulty-Aware Agentic Orchestration (DAAO), which dynamically generates query-specific workflows based on predicted difficulty, resulting in improved accuracy and inference efficiency across six benchmarks.
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.