Skip to content
conftrace_
2025 ICML ICML 2025

An Analysis for Reasoning Bias of Language Models with Small Initialization

Abstract

No abstract available. No abstract was captured for this paper when its proceedings were scraped - most often an older edition published without one online. How the corpus is built.

🌉 Interdisciplinary Bridge - Artificial Intelligence and Natural Language Processing

Authors

Junjie Yao, Zhongwang Zhang, Zhi-Qin John Xu

Topics

Machine Learning > Optimization & Theory > Learning Theory Natural Language Processing > Resources & Methods > Large Language Models Artificial Intelligence > Core AI > Large Language Models

Keywords

self-attention mechanism embedding space training dynamics reasoning task initialization scale memorization tasks
Download PDF