AdaReasoner: Adaptive Reasoning Enables More Flexible Thinking

Yanbo Wang (Mohamed bin Zayed University of Artificial Intelligence) · Yue Huang (Xiamen University) · Xiangliang Zhang (University of Notre Dame) · Xiaonan Luo (University of Notre Dame) · Kehan Guo (university of notre dame) · Yujun Zhou (University of Notre Dame) · Xiangqi Wang (University of Notre Dame)

adaptive reasoningadareasonerfactorized action spacefast convergencefew-shot guidellm-agnostic pluginpolicy model optimizationpretrained reward modelprompting approachesreinforcement learningsublinear policy gaptargeted exploration strategytask-specific optimalitytheoretical guarantees

LLMs often need effective configurations, like temperature and reasoning steps, to handle tasks requiring sophisticated reasoning and problem-solving, ranging from joke generation to mathematical reasoning. Existing prompting approaches usually adopt general-purpose, fixed configurations that work “well enough” across tasks but seldom achieve task-specific optimality. To address this gap, we introduce AdaReasoner, an LLM-agnostic plugin designed for any LLM to automate adaptive reasoning configurations for tasks requiring different types of thinking. AdaReasoner is trained using a reinforcement learning (RL) framework, combining a factorized action space with a targeted exploration strategy, along with a pretrained reward model to optimize the policy model for reasoning configurations with only a few-shot guide. AdaReasoner is backed by theoretical guarantees and experiments of fast convergence and a sublinear policy gap. Across six different LLMs and a variety of reasoning tasks, it consistently outperforms standard baselines, preserves out-of-distribution robustness, and yield gains on knowledge-intensive tasks through tailored prompts.