| title | Model Merging Scaling Laws in Large Language Models |
|---|---|
| authors | Yuanyi Wang, Yanggan Gu, Yiming Zhang, Qi Zhou, Zhaoyi Yan, Congkai Xie, Xinyao Wang, Jianbo Yuan, Hongxia Yang |
| date | 2026-05-11 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2509.24244 |
| affiliation | The Hong Kong Polytechnic University |
| upvotes | 42 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | We study empirical scaling laws for language model merging measured by cross-entropy. Despite its wide practical use, merging lacks a quantitative rule that predicts returns as we add experts or scale the model size. We identify a compact power law that links model size and expert number: the size-dependent floor decreases with model capacity, while the merging tail exhibits clear diminishing returns in the number of experts. The law holds in-domain and cross-domain, tightly fits measured curves across diverse architectures and methods (Average, TA, TIES, DARE), and explains two robust regularities: most gains arrive early, and variability shrinks as more experts are included. Building on this, we present a simple theory that explains why gains fall roughly as 1/k and links the floor and tail to properties of the base model and the diversity across domains. This law enables predictive planning: estimate how many experts are needed to reach a target loss, decide when to stop adding experts, and trade off scaling the base model versus adding experts under a fixed budget--turning merging from heuristic practice into a computationally efficient, planable alternative to multitask training. This suggests a scaling principle for distributed generative AI: predictable gains can be achieved by composing specialists, offering a complementary path toward AGI-level systems. |
cite
@misc{wang2026model,
title={Model Merging Scaling Laws in Large Language Models},
author={Yuanyi Wang and Yanggan Gu and Yiming Zhang and Qi Zhou and Zhaoyi Yan and Congkai Xie and Xinyao Wang and Jianbo Yuan and Hongxia Yang},
year={2026},
eprint={2509.24244},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2509.24244}
}