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 2601.15871cs.LGcs.AI
收藏

为什么在大模型中推理在训练后变得可分解

Why Inference in Large Models Becomes Decomposable After Training

Jidong Jin

AI总结:

研究发现大模型训练后推理系统结构非均匀且可分解,通过后训练统计标准和结构退火过程去除无效依赖,实现并行推理。

摘要:

在大规模AI模型中,推理通常在密集参数矩阵上进行,导致推理成本和系统复杂性随模型规模不可持续增长。这一限制并非由于模型容量不足,而是由于将训练后推理系统视为单一大型操作符,而忽略学习过程中形成的内部结构。我们证明在大模型中梯度更新事件高度局部化和选择性,训练后许多参数依赖在统计上与初始化分布无法区分。因此,训练后推理系统在结构上非均匀且本质上可分解。基于这一观察,我们引入了后训练统计标准和结构退火过程,以去除不支持的依赖并揭示稳定、独立的子结构。本工作建立了推理系统的后训练、模型无关的结构视图,并实现了无需修改模型功能或接口的结构化并行推理。

英文摘要:

Inference in large-scale AI models is typically performed on dense parameter matrices, leading to inference cost and system complexity that scale unsustainably with model size. This limitation does not arise from insufficient model capacity, but from treating post-training inference systems as monolithic operators while ignoring internal structures formed during learning. We show that gradient update events in large models are highly localized and selective, leaving many parameter dependencies statistically indistinguishable from their initialization distribution after training. As a result, post-training inference systems are structurally non-uniform and inherently decomposable. Based on this observation, we introduce a post-training statistical criterion and a structural annealing procedure that removes unsupported dependencies and reveals stable, independent substructures. This work establishes a post-training, model-agnostic structural view of inference systems and enables structured, parallel inference without modifying model functionality or interfaces.

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。

↑