Agentic Reasoning and Tool Integration for LLMs via Reinforcement Learning

TL;DR

This paper introduces ARTIST, a framework combining agentic reasoning, reinforcement learning, and tool integration to enhance LLMs' multi-step reasoning and interaction capabilities, leading to significant performance improvements.

Contribution

The paper presents a novel unified framework, ARTIST, that enables LLMs to autonomously decide when and how to use external tools through reinforcement learning, improving reasoning and problem-solving.

Findings

Up to 22% performance improvement over base models

Enhanced reasoning depth and tool use effectiveness

Strong results on mathematical and multi-turn function calling benchmarks

Abstract

Large language models (LLMs) have achieved remarkable progress in complex reasoning tasks, yet they remain fundamentally limited by their reliance on static internal knowledge and text-only reasoning. Real-world problem solving often demands dynamic, multi-step reasoning, adaptive decision making, and the ability to interact with external tools and environments. In this work, we introduce ARTIST (Agentic Reasoning and Tool Integration in Self-improving Transformers), a unified framework that tightly couples agentic reasoning, reinforcement learning, and tool integration for LLMs. ARTIST enables models to autonomously decide when, how, and which tools to invoke within multi-turn reasoning chains, leveraging outcome-based RL to learn robust strategies for tool use and environment interaction without requiring step-level supervision. Extensive experiments on mathematical reasoning and multi-turn function calling benchmarks show that ARTIST consistently outperforms state-of-the-art baselines, with up to 22% absolute improvement over base models and strong gains on the most challenging tasks. Detailed studies and metric analyses reveal that agentic RL training leads to deeper reasoning, more effective tool use, and higher-quality solutions. Our results establish agentic RL with tool integration as a powerful new frontier for robust, interpretable, and generalizable problem-solving in LLMs.