Mind the Tool Failures: Achieving Synergistic Tool Gains for Medical Agents

2026-05-26

AI Summary

This work addresses the safety risks in medical AI agents arising from instance-level failures of external tools, a challenge inadequately handled by existing approaches. To overcome the limitations of conventional task-level tool selection, the study formulates tool usage as an instance-level selection problem and introduces a novel instance-aware tool collaboration mechanism. Built upon the GRPO reinforcement learning framework, the method integrates a probability-risk-minimization reward, disagreement-aware collaborative learning, and an entropy-guided sampling strategy to correct erroneous consensus among tools at the instance level. Extensive experiments across two tasks and seven medical benchmarks demonstrate that the proposed approach substantially enhances system robustness and effectively narrows the risk gap between single-tool usage and an ideal instance-level selector.

Abstract

Medical AI agents increasingly use external tools for diagnosis, treatment recommendation, and evidence retrieval, yet most existing approaches assume that task-appropriate tools are reliable within their intended scope. This assumption is fragile in real clinical settings, where even relevant tools may fail on challenging instances and lead to unsafe downstream decisions. To address this issue, we study medical tool use under imperfect-tool settings to correct failure instances missed by individual tools. Instance-dependent failure patterns create a gap between the best fixed single tool and an ideal instance-wise selector, which we refer to as the Single-Oracle risk gap. The core challenge is that conventional task-level tool selection cannot realize this gap, as it is inherently bounded by the performance of the best single tool. Motivated by this observation, we therefore account for instance-level heterogeneity and formulate tool use as an instance-level selection problem. Particularly, we propose a GRPO-based reinforcement learning framework with rewards for probabilistic risk minimization and disagreement-aware synergy learning, which promotes instance-level correction of erroneous tool consensus. Furthermore, an entropy-guided sampling strategy is adopted to upweight high-disagreement instances, which provide stronger signals for learning instance-specific tool synergy. These two components complement each other in mitigating instance-level heterogeneity and improving tool synergy. Experiments on two tasks and seven medical benchmarks show that our method consistently achieves robust and stable improvements over a broad range of baselines, highlighting the importance of synergy-aware tool use for reliable medical agentic systems.

Problem

• tool failure
• medical AI agents
• instance-level heterogeneity
• tool reliability
• clinical decision-making

Innovation

• instance-level tool selection
• tool synergy
• GRPO reinforcement learning
• risk-aware medical AI
• disagreement-aware learning