展示，而非告知：将潜在推理转化为图像生成
Show, Don't Tell: Morphing Latent Reasoning into Image Generation
AI总结：
LatentMorph通过隐式潜在推理提升文本到图像生成的效率和效果，减少推理时间与标记消耗，同时提高与人类直觉的一致性。
摘要：
文本到图像（T2I）生成已经取得了显著进展，但现有方法往往缺乏在生成过程中动态推理和细化的能力——这是人类创造力的标志。当前的推理增强范式大多数依赖于显式思维过程，其中中间推理被解码为离散文本，在固定步骤中频繁进行图像解码和重编码，导致效率低下、信息丢失和认知不匹配。为弥合这一差距，我们引入了LatentMorph，一种新的框架，它无缝地将隐式潜在推理整合到T2I生成过程中。其核心是引入了四个轻量级组件：（i）一个摘要器，用于将中间生成状态总结为紧凑的视觉记忆；（ii）一个翻译器，用于将潜在思维转换为可操作的指导；（iii）一个形状器，用于动态地引导下一个图像标记的预测；（iv）一个通过强化学习训练的调用器，用于适应性地决定何时调用推理。通过在连续潜在空间中进行推理，LatentMorph避免了显式推理的瓶颈，使自我修正更加适应。广泛的实验表明，LatentMorph（I）在GenEval上比基模型Janus-Pro提高了16%，在T2I-CompBench上提高了25%；（II）在抽象推理任务如WISE和IPV-Txt上，比显式范式（如TwiG）高出15%和11%；（III）同时减少了推理时间44%和标记消耗51%；（IV）在推理调用上表现出71%的认知与人类直觉的一致性。
英文摘要：
Text-to-image (T2I) generation has achieved remarkable progress, yet existing methods often lack the ability to dynamically reason and refine during generation--a hallmark of human creativity. Current reasoning-augmented paradigms most rely on explicit thought processes, where intermediate reasoning is decoded into discrete text at fixed steps with frequent image decoding and re-encoding, leading to inefficiencies, information loss, and cognitive mismatches. To bridge this gap, we introduce LatentMorph, a novel framework that seamlessly integrates implicit latent reasoning into the T2I generation process. At its core, LatentMorph introduces four lightweight components: (i) a condenser for summarizing intermediate generation states into compact visual memory, (ii) a translator for converting latent thoughts into actionable guidance, (iii) a shaper for dynamically steering next image token predictions, and (iv) an RL-trained invoker for adaptively determining when to invoke reasoning. By performing reasoning entirely in continuous latent spaces, LatentMorph avoids the bottlenecks of explicit reasoning and enables more adaptive self-refinement. Extensive experiments demonstrate that LatentMorph (I) enhances the base model Janus-Pro by $16\%$ on GenEval and $25\%$ on T2I-CompBench; (II) outperforms explicit paradigms (e.g., TwiG) by $15\%$ and $11\%$ on abstract reasoning tasks like WISE and IPV-Txt, (III) while reducing inference time by $44\%$ and token consumption by $51\%$; and (IV) exhibits $71\%$ cognitive alignment with human intuition on reasoning invocation.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
