@inproceedings{li-etal-2026-wisca,
    title = "{WISCA}: A Lightweight Model Transition Method to Improve {LLM} Training via Weight Scaling",
    author = "Li, Jiacheng  and
      Tan, Jianchao  and
      Yang, Zhidong  and
      Sun, Pingwei  and
      Huo, Feiye  and
      Qin, Jiayu  and
      Zhang, Xiangyu  and
      He, Maoxin  and
      Tan, Guangming  and
      Jia, Weile  and
      Cai, Xunliang  and
      Zhao, Tong",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.79/",
    pages = "1589--1601",
    ISBN = "979-8-89176-395-1",
    abstract = "Transformer architecture gradually dominates the LLM field. Recent advances in training optimization for Transformer-based large language models (LLMs) primarily focus on architectural modifications or optimizer adjustments. However, these approaches lack systematic optimization of weight patterns during training. Weight pattern refers to the distribution and relative magnitudes of weight parameters in a neural network. To address this issue, we propose a Weight Scaling method called WISCA to enhance training efficiency and model quality by strategically improving neural network weight patterns{---}without changing network structures. By rescaling weights while preserving model outputs, WISCA indirectly optimizes the model{'}s training trajectory. Experiments demonstrate that WISCA significantly improves convergence quality (measured by generalization capability and loss reduction), particularly in LLMs with Grouped Query Attention (GQA) architectures and LoRA fine-tuning tasks. Empirical results show 5.6{\%} average improvement on zero-shot validation tasks and 2.12{\%} average reduction in training perplexity across multiple architectures."
}