arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.11965cs.LGcs.AI
收藏

面向流形的时域泛化用于大语言模型

Parameter-Efficient Temporal Domain Generalization

The University of Tokyo(东京大学);LocationMind;Emory University(埃默里大学);Jilin University(吉林大学)

Zekun Cai, Xinyuan Song, Yiheng Yao, Xuan Song, Ryosuke Shibasaki, Liang Zhao

AI总结:

本文提出MaT-LoRA,通过在低秩适应子空间内约束时间更新到共享低维流形,实现大语言模型的时间域泛化,提升模型的时间建模效率与泛化能力。

摘要:

在大语言模型(LLMs)的实际部署中,时间分布偏移普遍存在,数据会持续随时间变化。虽然时间域泛化(TDG)旨在建模这种结构化演变,但现有方法在完整参数空间中建模模型适应,这种形式对于现代LLMs来说计算上不可行。本文提出了一种在参数高效微调下的TDG几何重新表述。我们证明,模型演变下的低维时间结构可以在参数高效重参数化下得以保留,从而在不操作于环境参数空间的情况下实现时间建模。基于这一原理,我们提出了面向流形的时间LoRA(MaT-LoRA),它将时间更新限制在低秩适应子空间内的共享低维流形中,并通过结构化时间核心建模其演变。这种重参数化显著降低了时间建模的复杂性,同时保留了表达能力。在合成和真实世界数据集上进行了广泛的实验,包括科学文档、新闻出版商和评论评分,结果表明,MaT-LoRA在LLMs中实现了优越的时间泛化性能,并具有实用的可扩展性。

英文摘要:

Temporal domain generalization (TDG) learns from a sequence of observed domains to predict models for unseen future domains. Existing parameter-dynamics methods become prohibitively expensive at modern backbone scales because their temporal encoders and generators operate on the entire adapted parameter state. Replacing full-model updates with LoRA does not remove this failure mode: the dynamics model must still encode and reconstruct every adapter parameter, making the auxiliary model itself prohibitively large for modern backbones. We propose Low-rank Time-varying Core (LoTiC), a parameter-efficient TDG method that decomposes temporal updates into shared low-rank bases and a compact core. The bases define a stable update subspace, while only the core evolves. LoTiC therefore learns temporal dynamics in an r*r space instead of tracking the full adapter. We justify this factorization through the geometry of parameter increments and support continuous, recurrent, and time-conditioned core models. Experiments span six controlled and real-world benchmarks, three data modalities, classification and generation, and backbones from 66M to 8B parameters. LoTiC ranks first in 14 of 15 settings, including near-perfect future-domain accuracy with an 8B backbone on Rotating 2-Moons. It adds only 0.13--0.51% trainable parameters beyond LoRA and trains 1.8--7.1$\times$ faster than parameter-dynamics baselines. These results establish that temporal parameter dynamics can remain both learnable and efficient as pretrained backbones scale.

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。

↑