arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.06993cs.LGcs.CL
收藏

吸引子补丁网络:通过路由低秩补丁专家减少灾难性遗忘

Attractor Patch Networks: Reducing Catastrophic Forgetting with Routed Low-Rank Patch Experts

Shashank

Shashank

AI总结:

吸引子补丁网络通过路由低秩补丁专家减少灾难性遗忘,提升持续学习的适应能力。

摘要:

Transformer在语言建模准确性方面表现强劲,但其位置-wise前馈网络(FFN)是密集的、全局共享的,并且通常端到端更新。这些特性产生了两个实际矛盾。首先,密集的FFN在每个token上消耗相同的计算资源,无论上下文如何,即使语言表现出高度聚类的上下文结构,也分配容量均匀。其次,持续学习,即在数据流中更新模型,往往会产生干扰,因为小的更新会触碰广泛共享的权重。我们提出了吸引子补丁网络(APN),一种可以无缝替换Transformer FFN的架构。APN是一组补丁专家的银行。一个相似性路由器通过将token表示与学习到的原型匹配,为每个token选择一个小的top-k补丁集。每个选定的补丁会根据一个紧凑的代码发射一个低秩残差更新。该架构在保持标准Transformer接口的同时,实现了条件性的、上下文专门化的非线性转换。本文专注于APN作为架构基本元素。我们正式化了APN,分析了其作为分段低秩残差函数类的表达能力,并推导了简单的干扰和稳定性论点,使APN自然兼容持续学习。在字符级语言建模实验中,APN在保持竞争力的困惑度(4.57 vs 4.32 PPL)的同时,显著提升了持续适应能力:当适应到移位领域时,APN在原始领域中的保留能力比密集FFN基线提高了2.6倍(11.1 vs 29.4 PPL),在新领域中的适应能力提高了2.8倍(6.4 vs 17.8 PPL)。

英文摘要:

Transformers achieve strong language modeling accuracy, yet their position-wise feed-forward networks (FFNs) are dense, globally shared, and typically updated end to end. These properties create two practical tensions. First, dense FFNs spend the same compute on every token regardless of context, and they allocate capacity uniformly even when language exhibits highly clustered context structure. Second, continual learning, in the sense of updating the model while serving a data stream, often produces interference because a small update touches broadly shared weights. We propose Attractor Patch Networks (APN), a plug-compatible replacement for the Transformer FFN. APN is a bank of patch experts. A similarity router selects a small top-k set of patches for each token by matching the token representation to learned prototypes. Each selected patch emits a low-rank residual update conditioned on a compact code. The architecture yields conditional, context-specialized nonlinear transformations while preserving the standard Transformer interface. This paper focuses on APN as an architectural primitive. We formalize APN, analyze its expressivity as a piecewise low-rank residual function class, and derive simple interference and stability arguments that make APN naturally compatible with continual learning. In experiments on character-level language modeling, APN achieves competitive perplexity (4.57 vs 4.32 PPL) while enabling dramatically better continual adaptation: when adapting to a shifted domain, APN achieves 2.6 times better retention (11.1 vs 29.4 PPL on the original domain) and 2.8 times better adaptation (6.4 vs 17.8 PPL on the new domain) compared to global fine-tuning of a dense FFN baseline.

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。

↑