paper-with-me

홈 › Papers

An Analysis for Reasoning Bias of Language Models with Small Initialization

2025-02-05 · Junjie Yao, Zhongwang Zhang, Zhi-Qin John Xu

Transformer-based Large Language Models (LLMs) have revolutionized Natural Language Processing by demonstrating exceptional performance across diverse tasks. This study investigates the impact of the parameter initialization scale on the training behavior and task preferences of LLMs. We discover that smaller initialization scales encourage models to favor reasoning tasks, whereas larger initialization scales lead to a preference for memorization tasks. We validate this reasoning bias via real datasets and meticulously designed anchor functions. Further analysis of initial training dynamics suggests that specific model components, particularly the embedding space and self-attention mechanisms, play pivotal roles in shaping these learning biases. We provide a theoretical framework from the perspective of model training dynamics to explain these phenomena. Additionally, experiments on real-world language tasks corroborate our theoretical insights. This work enhances our understanding of how initialization strategies influence LLM performance on reasoning tasks and offers valuable guidelines for training models.

📄 PDF Abstract BibTeX arXiv:2502.04375

Code (0)

등록된 구현이 없습니다.

Tasks

Memorization

Similar Papers 제목 키워드 기반

Small Initialization Matters for Large Language Models

2026-06-16 · Liangkai Hang, Junjie Yao, Zhiyu Li, Feiyu Xiong 외 arxiv

Large language models provide a tractable system for asking how intelligence itself emerges, rather than only how LLMs can be engineered. Although progress is usually attributed to scale, data and architecture, we show t…

Implicit Bias of Linear RNNs

2021-01-19 · Melikasadat Emami, Mojtaba Sahraee-Ardakan, Parthe Pandit, Sundeep Rangan 외

Contemporary wisdom based on empirical studies suggests that standard recurrent neural networks (RNNs) do not perform well on tasks requiring long-term memory. However, precise reasoning for this behavior is still unknow…

Small random initialization is akin to spectral learning: Optimization and generalization guarantees for overparameterized low-rank matrix reconstruction

2021-06-28 · NeurIPS 2021 12 · Dominik Stöger, Mahdi Soltanolkotabi

Recently there has been significant theoretical progress on understanding the convergence and generalization of gradient-based methods on nonconvex losses with overparameterized models. Nevertheless, many aspects of opti…

From Condensation to Rank Collapse: A Two-Stage Analysis of Transformer Training Dynamics

2025-10-08 · Zheng-An Chen, Tao Luo arxiv

Although transformer-based models have shown exceptional empirical performance, the fundamental principles governing their training dynamics are inadequately characterized beyond configuration-specific studies. Inspired …

Convolutional Initialization for Data-Efficient Vision Transformers

2024-01-23 · Jianqiao Zheng, Xueqian Li, Simon Lucey

Training vision transformer networks on small datasets poses challenges. In contrast, convolutional neural networks (CNNs) can achieve state-of-the-art performance by leveraging their architectural inductive bias. In thi…

Inductive Bias