SPA-Cache: 扩散语言模型中的自适应缓存奇异代理
SPA-Cache: Singular Proxies for Adaptive Caching in Diffusion Language Models
AI总结：
针对扩散语言模型因非因果特性无法使用标准KV缓存导致计算开销大的问题，提出SPA-Cache方法，通过低维奇异代理识别关键令牌并自适应分配缓存预算，实现高达8倍吞吐量提升和2-4倍加速。
摘要：
尽管扩散语言模型（DLM）为自回归范式提供了一种灵活、任意顺序的替代方案，但其非因果特性排除了标准的KV缓存，迫使在每个解码步骤进行昂贵的隐藏状态重新计算。现有的DLM缓存方法通过选择性隐藏状态更新来降低这一成本；然而，它们仍然受限于（i）昂贵的逐令牌更新识别启发式方法和（ii）僵化的统一预算分配，未能考虑异构的隐藏状态动态。为了解决这些挑战，我们提出了SPA-Cache，它在DLM缓存中联合优化了更新识别和预算分配。首先，我们推导出一个低维奇异代理，能够在低维子空间中识别更新关键令牌，大幅降低更新识别的开销。其次，我们引入一种自适应策略，在不降低生成质量的情况下，为稳定层分配更少的更新。这些贡献共同显著提高了DLM的效率，相比原始解码实现了高达8倍的吞吐量提升，相比现有缓存基线实现了2-4倍的加速。
英文摘要：
While Diffusion Language Models (DLMs) offer a flexible, arbitrary-order alternative to the autoregressive paradigm, their non-causal nature precludes standard KV caching, forcing costly hidden state recomputation at every decoding step. Existing DLM caching approaches reduce this cost by selective hidden state updates; however, they are still limited by (i) costly token-wise update identification heuristics and (ii) rigid, uniform budget allocation that fails to account for heterogeneous hidden state dynamics. To address these challenges, we present SPA-Cache that jointly optimizes update identification and budget allocation in DLM cache. First, we derive a low-dimensional singular proxy that enables the identification of update-critical tokens in a low-dimensional subspace, substantially reducing the overhead of update identification. Second, we introduce an adaptive strategy that allocates fewer updates to stable layers without degrading generation quality. Together, these contributions significantly improve the efficiency of DLMs, yielding up to an $8\times$ throughput improvement over vanilla decoding and a $2$--$4\times$ speedup over existing caching baselines.
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。
