EASy: Towards Efficient LLM-Based Agentic System
Junnan Liu, Linhao Luo, Thuy-Trang Vu, Gholamreza Haffari
Why It Matters
What makes this one worth your time
This work is relevant for AI engineers and researchers looking to enhance the efficiency of LLM-based systems, particularly in scenarios with practical constraints on computational resources.
EASy optimizes LLM-based agentic systems for both performance and efficiency using a novel reinforcement learning approach.
Summary
The paper introduces EASy, a trainable agentic framework that optimizes task performance and computational efficiency using reinforcement learning. It features an LLM-based orchestrator that coordinates tasks by considering executor capabilities and costs, and employs a milestone-plan-act workflow to manage complex tasks. The framework is trained using a tree-structured rollout procedure with multi-component rewards, and it demonstrates improved performance-efficiency trade-offs in various benchmarks.
Key contributions
- Development of a trainable agentic framework using reinforcement learning for efficiency optimization.
- Introduction of a milestone-plan-act workflow for task decomposition and execution.
- Implementation of a tree-structured rollout procedure for orchestrator training.
Notable insights
- The use of a milestone-plan-act workflow to decompose tasks and adapt execution plans based on intermediate outcomes.
- Incorporating executor capability and cost profiles into the coordination process for better efficiency.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2608.04588v1 Announce Type: cross Abstract: Agentic systems have emerged as a promising paradigm for solving complex tasks by coordinating specialized LLM-based agents. However, most existing systems primarily optimize task success while giving limited consideration to execution efficiency under practical constraints such as executor capability and computational cost. Existing router-based methods have limited ability to reason over rich, evolving task contexts, multi-step dependencies, and intermediate execution feedback, and often generalize poorly to unseen executors. We propose EASy, a trainable agentic framework that jointly optimizes task performance and computational efficiency through reinforcement learning. EASy equips an LLM-based orchestrator with explicit knowledge of the capability and cost profiles of heterogeneous executors, enabling context-sensitive coordination beyond performance-only routing. It further introduces a milestone-plan-act workflow that decomposes complex tasks into manageable milestones, constructs dependency-aware execution graphs, assigns suitable executors, and parallelizes independent steps while adapting subsequent decisions to intermediate outcomes. To train the orchestrator, we develop a tree-structured rollout procedure that explores alternative milestone decompositions and execution plans, together with multi-component rewards that capture task correctness, execution efficiency, and trajectory completeness. Extensive experiments on mathematical reasoning, embodied decision-making, and deep research benchmarks show that EASy consistently achieves stronger performance-efficiency trade-offs than strong agentic baselines.