森林在树之前:用于高效视觉推理的潜在叠加
Forest Before Trees: Latent Superposition for Efficient Visual Reasoning
AI总结:
本文提出Laser方法,通过动态窗口对齐学习实现视觉推理的'森林在树之前'认知层次,提升模型在保持全局特征叠加的同时,实现高效且可解释的视觉推理。
摘要:
尽管Chain-of-Thought使大视觉-语言模型具备多步推理能力,但显式的文本推理却因信息带宽瓶颈而受限,连续视觉细节在离散标记化过程中被丢弃。最近的潜在推理方法试图解决这一挑战,但往往因刚性的自回归目标而陷入提前语义坍缩。本文提出Laser,一种新的范式,通过动态窗口对齐学习(DWAL)重新表述视觉推理。与强制点对点预测不同,Laser将潜在状态与未来语义的动态有效性窗口对齐。这种机制强制形成'森林在树之前'的认知层次,使模型在聚焦局部细节之前,能够保持全局特征的概率叠加。关键的是,Laser通过可解码轨迹保持可解释性,同时通过自精炼叠加稳定无约束学习。在6个基准上的广泛实验表明,Laser在潜在推理方法中实现了最先进的性能,平均比强基线Monet高出5.03%。值得注意的是,它在极端高效性方面实现了这些收益,推理标记数减少超过97%,同时在分布外领域表现出鲁棒的泛化能力。
英文摘要:
While Chain-of-Thought empowers Large Vision-Language Models with multi-step reasoning, explicit textual rationales suffer from an information bandwidth bottleneck, where continuous visual details are discarded during discrete tokenization. Recent latent reasoning methods attempt to address this challenge, but often fall prey to premature semantic collapse due to rigid autoregressive objectives. In this paper, we propose Laser, a novel paradigm that reformulates visual deduction via Dynamic Windowed Alignment Learning (DWAL). Instead of forcing a point-wise prediction, Laser aligns the latent state with a dynamic validity window of future semantics. This mechanism enforces a "Forest-before-Trees" cognitive hierarchy, enabling the model to maintain a probabilistic superposition of global features before narrowing down to local details. Crucially, Laser maintains interpretability via decodable trajectories while stabilizing unconstrained learning via Self-Refined Superposition. Extensive experiments on 6 benchmarks demonstrate that Laser achieves state-of-the-art performance among latent reasoning methods, surpassing the strong baseline Monet by 5.03% on average. Notably, it achieves these gains with extreme efficiency, reducing inference tokens by more than 97%, while demonstrating robust generalization to out-of-distribution domains.
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。