EASy: Towards Efficient LLM-Based Agentic System

2026-08-05

AI Summary

This work addresses the tendency of existing large language model agents to overlook execution efficiency in complex tasks, often failing to balance performance with practical constraints. The authors propose EASy, a framework that jointly optimizes task performance and computational efficiency through reinforcement learning. EASy decomposes tasks via a milestone-planning-execution pipeline, constructs a dependency-aware execution graph, and dynamically schedules heterogeneous executors to process independent subtasks in parallel. A trainable coordinator integrates executor capabilities with cost awareness, leveraging tree-based backtracking exploration and a multi-dimensional reward mechanism to significantly enhance generalization and collaborative efficiency on unseen executors. Experiments demonstrate that EASy achieves a superior trade-off between task success rate and computational efficiency across benchmarks in mathematical reasoning, embodied decision-making, and deep research, substantially outperforming current approaches.

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.

Problem

• agentic system
• execution efficiency
• computational cost
• task decomposition
• executor generalization

Innovation

• agentic system
• reinforcement learning
• execution efficiency
• milestone decomposition
• heterogeneous executors