结构-残差归纳偏助语言模型预训练研究:BBC 与 CC 方法解析
Structure--Residual Inductive Biases for Language Model Pretraining
AI 导读
一项研究提出“结构-残差”归纳偏置原则,让语言模型预训练中的表示显式分离结构与实现细节,包含 Boundary Bottleneck Concept(BBC)和 Continuous Concept(CC)两种实现。
正文 · 原文
Abstract:Next-token prediction is relied upon to learn both predictive behavior and internal representations. We ask whether representation structure itself should be an explicit inductive bias. Motivated by the two-part coding view of Kolmogorov complexity, we study a structure--residual principle in which a representation separates structure from realization detail. We instantiate this principle in two complementary ways. Boundary Bottleneck Concept (BBC) routes information from token segments through boundary carriers while preserving local autoregressive computation. Continuous Concept (CC) regularizes representation geometry and induces a spectral decomposition into low-frequency structure and high-frequency residual. We derive coding bounds for both views and show that multiplicity-corrected boundary attention becomes exact as the within-segment path energy vanishes. Empirically, BBC and CC exhibit strikingly different depth preferences: BBC works best in early layers, whereas CC is most effective in later layers. At 300M training tokens, combining BBC and CC improves NLL by 1.44\% on GPT-2 Medium. These gains persist with training runs up to 20B tokens. Representation diagnostics show that BBC increases the relative predictive importance of boundary states, while CC reduces spectral residual energy. Together, these results suggest that structure--residual biases can improve language-model pretraining beyond next-token prediction. Code: this https URL
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.23197 [cs.CL] |
| (or arXiv:2611.23197v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.23197 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hai Huang [view email]
[v1]
Tue, 17 Nov 2026 18:38:27 UTC (265 KB)
来源:arXiv cs.AI · arxiv.org