AdaTIR: Adaptive Tool-Integrated Reasoning via Difficulty-Aware Policy Optimization

TL;DR

AdaTIR introduces a dynamic, difficulty-aware reasoning framework for LLMs that reduces unnecessary tool use and internalizes simple reasoning, significantly improving efficiency and accuracy.

Contribution

It proposes AdaTIR, a novel framework that adaptively manages tool invocation based on task difficulty, and introduces Clipped Advantage Shaping to address reward sign reversal issues.

Findings

Reduces tool calls by up to 97.6% on simple tasks

Maintains or improves accuracy with fewer tool invocations

Outperforms baselines by 4.8% on AIME 2024 without tool access

Abstract

Tool-Integrated Reasoning (TIR) has significantly enhanced the capabilities of Large Language Models (LLMs), yet current agents tend to exhibit cognitive offloading, redundantly invoking external tools even for simple tasks. In this paper, we suggest that true agentic intelligence requires not just tool invocation, but the adaptive wisdom to discern when to use them. We propose AdaTIR, a framework that shifts the paradigm from static tool invocation to difficulty-aware reasoning internalization. By introducing a difficulty-aware efficiency reward, AdaTIR dynamically adjusts tool budgets based on task complexity--internalizing reasoning for simple tasks while selectively invoking tools for complex tasks. Furthermore, we identify a sign reversal problem where tool penalties outweigh correctness rewards, mistakenly penalizing correct rollouts with negative advantages. To resolve this, we propose Clipped Advantage Shaping (CAS), which ensures that correctness remains the primary objective while using efficiency as a secondary constraint. Empirical results demonstrate that AdaTIR reduces tool calls by up to 97.6% on simple tasks and 28.2% on complex challenges while maintaining or enhancing accuracy. Notably, AdaTIR successfully internalizes reasoning, outperforming baselines by 4.8% on AIME 2024 even when tool access is strictly disabled.