← Search

Zhaoxin Huan

2 accepted papers

2025

BOSE: A Systematic Evaluation Method Optimized for Base Models

ACL 2025finding

This paper poses two critical issues in evaluating base models (without post-training): (1) Unstable evaluation during training: in the early stages of pre-training, the models lack the capability to answer questions as required, leading to unstable evaluation results. This instability makes it diff…

2024

Rethinking Memory and Communication Costs for Efficient Data Parallel Training of Large Language Models

NeurIPS 2024poster

Recently, various strategies for distributed training of large language models (LLMs) have been proposed. By categorizing them into basic strategies and composite strategies, we have discovered that existing basic strategies provide limited options in specific scenarios, leaving considerable room fo…

Cited by 0SourcePDFScholar