AdaReasoner: Dynamic Tool Orchestration for Iterative Visual Reasoning

TL;DR

AdaReasoner tackles visual reasoning beyond innate capabilities by teaching models to use external tools as a general reasoning skill. It combines a scalable data curation pipeline, Tool-GRPO reinforcement learning for long-horizon tool sequencing, and adaptive learning to generalize to unseen tools and tasks. Empirically, it demonstrates strong tool-adaptive behavior, high generalization to new tools, and state-of-the-art performance on multiple benchmarks, including surpassing GPT-5 on several tasks with a 7B model. The work suggests shifting the bottleneck from model scale to tool utility, enabling smaller open-source models to reach leading performance when equipped with a capable tool suite.

Abstract

When humans face problems beyond their immediate capabilities, they rely on tools, providing a promising paradigm for improving visual reasoning in multimodal large language models (MLLMs). Effective reasoning, therefore, hinges on knowing which tools to use, when to invoke them, and how to compose them over multiple steps, even when faced with new tools or new tasks. We introduce \textbf{AdaReasoner}, a family of multimodal models that learn tool use as a general reasoning skill rather than as tool-specific or explicitly supervised behavior. AdaReasoner is enabled by (i) a scalable data curation pipeline exposing models to long-horizon, multi-step tool interactions; (ii) Tool-GRPO, a reinforcement learning algorithm that optimizes tool selection and sequencing based on end-task success; and (iii) an adaptive learning mechanism that dynamically regulates tool usage. Together, these components allow models to infer tool utility from task context and intermediate outcomes, enabling coordination of multiple tools and generalization to unseen tools. Empirically, AdaReasoner exhibits strong tool-adaptive and generalization behaviors: it autonomously adopts beneficial tools, suppresses irrelevant ones, and adjusts tool usage frequency based on task demands, despite never being explicitly trained to do so. These capabilities translate into state-of-the-art performance across challenging benchmarks, improving the 7B base model by +24.9\% on average and surpassing strong proprietary systems such as GPT-5 on multiple tasks, including VSP and Jigsaw.

Figures (11)

• Figure 1: AdaReasoner performs adaptive and generalized tool-using.
• Figure 2: An overview of our AdaReasoner framework. The pipeline consists of two main stages: (a) the Tool Cold Start (TC) phase, where trajectories are carefully constructed to support multi-turn reasoning; and (b) the Tool GRPO (TG) phase, where the policy is further refined via reinforcement learning guided by our adaptive, multi-turn reward. In addition, the Adaptive Learning method (c) can be applied throughout both the TC and TG stages, enabling improved generalization across tasks and tool configurations.
• Figure 3: Trend for tool calling frequencies for AStar, Point, and Draw2DPath during RL. The model is optimized on VSP Verification (cool-color) and VSP Navigation (warm-color) tasks.
• Figure 4: Our AdaReasoner-7B demonstrates advanced capabilities for multi-turn, tool-assisted reasoning and reflection, enabling it to achieve performance that is on par with, or even superior to, state-of-the-art closed-source models.
• Figure 5: An example of a multi-turn cold-start trajectory for the VSP task.