Ares: Adaptive Reasoning Effort Selection for Efficient LLM Agents

TL;DR

Ares is a dynamic framework that intelligently adjusts reasoning effort per step in LLM agents, reducing inference costs significantly while maintaining high task success rates across various complex multi-step tasks.

Contribution

We introduce Ares, a novel lightweight router that predicts optimal reasoning effort levels for each step, enabling adaptive and cost-effective LLM agent reasoning strategies.

Findings

Reduces reasoning token usage by up to 52.7%

Maintains high task success rates with minimal performance loss

Effective across diverse multi-step agent tasks

Abstract

Modern agents powered by thinking LLMs achieve high accuracy through long chain-of-thought reasoning but incur substantial inference costs. While many LLMs now support configurable reasoning levels (e.g., high/medium/low), static strategies are often ineffective: using low-effort modes at every step leads to significant performance degradation, while random selection fails to preserve accuracy or provide meaningful cost reduction. However, agents should reserve high reasoning effort for difficult steps like navigating complex website structures, while using lower-effort modes for simpler steps like opening a target URL. In this paper, we propose Ares, a framework for per-step dynamic reasoning effort selection tailored for multi-step agent tasks. Ares employs a lightweight router to predict the lowest appropriate reasoning level for each step based on the interaction history. To train this router, we develop a data generation pipeline that identifies the minimum reasoning effort required for successful step completion. We then fine-tune the router to predict these levels, enabling plug-and-play integration for any LLM agents. We evaluate Ares on a diverse set of agent tasks, including TAU-Bench for tool use agents, BrowseComp-Plus for deep-research agents, and WebArena for web agents. Experimental results show that Ares reduces reasoning token usage by up to 52.7% compared to fixed high-effort reasoning, while introducing minimal degradation in task success rates.