Interleaved reasoning paradigms enhance Multimodal Large Language Models (MLLMs) with visual feedback but are hindered by the prohibitive computational cost of repeatedly re-encoding pixel-dense images. A promising alternative, latent visual reasoning, circumvents this bottleneck yet currently forces a critical trade-off: methods either sacrifice precise perceptual modeling by over-compressing features or fail to model dynamic problems due to static, non-interleaved structures. We introduce Interleaved Latent Visual Reasoning (ILVR), a framework that unifies dynamic state evolution with precise perceptual modeling. ILVR interleaves textual generation with latent visual representations that act as specific, evolving cues for subsequent reasoning. To enable this, we employ a self-supervision strategy where a Momentum Teacher Model selectively distills relevant features from helper images into sparse supervision targets. This adaptive selection mechanism guides the model to autonomously generate context-aware visual signals. Extensive experiments on multimodal reasoning benchmarks demonstrate that ILVR significantly outperforms existing approaches, effectively bridging the gap between fine-grained perception and sequential multimodal reasoning.

翻译：交错式推理范式通过视觉反馈增强多模态大语言模型（MLLMs），但受限于重复编码像素密集图像的高昂计算成本。一种有前景的替代方案——潜在视觉推理——规避了这一瓶颈，但当前面临关键权衡：现有方法要么因过度压缩特征而牺牲精确感知建模，要么因静态非交错结构而无法建模动态问题。本文提出交错式潜在视觉推理（ILVR），该框架将动态状态演化与精确感知建模相统一。ILVR将文本生成与潜在视觉表征交错结合，这些表征作为后续推理的具体演化线索。为实现此目标，我们采用自监督策略，其中动量教师模型从辅助图像中选择性蒸馏相关特征至稀疏监督目标。这种自适应选择机制引导模型自主生成上下文感知的视觉信号。在多模态推理基准上的大量实验表明，ILVR显著优于现有方法，有效弥合了细粒度感知与序列多模态推理之间的鸿沟。