arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.19169cs.LGcs.AIcs.MSmath.PR
收藏

虚拟参数锐化:用于推理时间推理增强的动态低秩扰动

Virtual Parameter Sharpening: Dynamic Low-Rank Perturbations for Inference-Time Reasoning Enhancement

Independent Researcher(独立研究者)

Saba Kublashvili

AI总结:

虚拟参数锐化通过动态低秩扰动提升大语言模型推理能力,实现测试时间适应与高效推理增强。

摘要:

我介绍了虚拟参数锐化(VPS),一种推理时间技术,该技术通过动态、激活条件的低秩扰动来增强冻结的Transformer线性层。与参数高效微调方法如LoRA不同,后者学习静态的低秩适配器,VPS会根据批量激活统计信息和可选梯度信号即时构建其扰动因子,从而在测试时间进行适应而无需持续参数更新。扰动形式为Delta W = gamma * W^T V U^T W,其中选择矩阵U和V通过稀疏激活引导选择或Syylvester耦合回归构建。我们对扰动的谱性质进行了理论分析,并描述了一个自适应策略系统,该系统根据激活能量和令牌级熵调节扰动大小。该系统结合了多目标验证与迭代优化,用于具有真实监督的任务。我们提出了完整的算法框架,分析了其数学基础,并讨论了激活条件计算如何增强大型语言模型的推理能力。实现和实验代码可在https://github.com/Saba-Kublashvili/vps-virtual-parameter-synthesis 上获得。

英文摘要:

I introduce Virtual Parameter Sharpening (VPS), an inference-time technique that augments frozen transformer linear layers with dynamic, activation-conditioned low-rank perturbations. Unlike parameter-efficient fine-tuning methods such as LoRA, which learn static low-rank adapters, VPS constructs its perturbation factors on the fly from batch activation statistics and optional gradient signals, enabling test-time adaptation without persistent parameter updates. The perturbation takes the form Delta W = gamma * W^T V U^T W, where selector matrices U and V are constructed via sparse activation-guided selection or Sylvester-coupled regression. We provide a theoretical analysis of the perturbation's spectral properties and describe an adaptive policy system that modulates perturbation magnitude based on activation energy and token-level entropy. This system incorporates multi-objective verification with iterative refinement for tasks with ground-truth supervision. We present the complete algorithmic framework, analyze its mathematical foundations, and discuss the mechanisms by which activation-conditioned computation may enhance reasoning capabilities in large language models. Implementation and experimental code are available at https://github.com/Saba-Kublashvili/vps-virtual-parameter-synthesis .

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。

↑