AI 中文总结
DiffuseAgent-MI是基于分布、集成工具的自进化视觉推理智能体，通过KL最小能量模型与轨迹级验证器提升忠实性，在多数据集上准确率及相关指标表现优于现有模型。
AI 中文摘要
集成工具的视觉-语言智能体在组合式多步骤视觉推理任务中取得了显著进展，但其输出常存在不忠实问题：所陈述的推理路径与实际生成答案的计算过程不一致，削弱了安全关键应用中的可靠性。本文提出DiffuseAgent-MI，这是一种自进化智能体，其感知基础由基于特征单元的KL最小能量模型控制，提供了视觉机制可解释性的分布视角。该智能体学习能量景观，软性约束生成样本位于所选可解释单元条件下的原生先验附近，缩小解释与内部表征间的差距。随后，验证器提供轨迹级忠实性奖励，当验证器标记某步骤不忠实，修复分支会重新调整能量条件。在GeoQA、SciVis、VQA-v2及内部多模态推理数据集上，DiffuseAgent-MI较现有自进化智能体提升了最高5.1个百分点的准确率，同时将互信息忠实性与人类可解释性一致性提升一倍以上。分析表明，能量项与验证器互补：前者保证分布级忠实性，后者保证轨迹级忠实性，仅二者结合可同时缩小两类差距。
英文摘要
Tool-integrated vision-language agents have made remarkable progress on compositional and multi-step visual reasoning. Yet their outputs frequently exhibit unfaithfulness: the stated reasoning path diverges from the computation that actually produced the answer, undermining reliability in safety-critical applications. We present DiffuseAgent-MI, a self-evolving agent whose perceptual grounding is governed by a KL-minimal energy model over feature units, providing a distributional view of visual mechanistic interpretability. The agent learns an energy landscape that softly constrains generated samples to lie near the native prior conditioned on the chosen interpretable unit, closing the gap between the explanation and the internal representation. A verifier then supplies trajectory-level faithfulness rewards, and a repair branch re-conditions the energy when the verifier flags an unfaithful step. On GeoQA, SciVis, VQA-v2 and an in-house multimodal reasoning set, DiffuseAgent-MI improves accuracy by up to 5.1 points over prior self-evolving agents while more than doubling mutual-information faithfulness and human-interpretability agreement. Our analysis shows the energy term and the verifier are complementary: the former guarantees distributional faithfulness, the latter trajectory-level faithfulness, and only their combination closes both gaps.
Comments11 pages, 9 figures, accepted by ICML 2026 manitrack
