English

Weaver: Foundation Models for Creative Writing

Computation and Language 2024-01-31 v1 Artificial Intelligence Machine Learning

Abstract

This work introduces Weaver, our first family of large language models (LLMs) dedicated to content creation. Weaver is pre-trained on a carefully selected corpus that focuses on improving the writing capabilities of large language models. We then fine-tune Weaver for creative and professional writing purposes and align it to the preference of professional writers using a suit of novel methods for instruction data synthesis and LLM alignment, making it able to produce more human-like texts and follow more diverse instructions for content creation. The Weaver family consists of models of Weaver Mini (1.8B), Weaver Base (6B), Weaver Pro (14B), and Weaver Ultra (34B) sizes, suitable for different applications and can be dynamically dispatched by a routing agent according to query complexity to balance response quality and computation cost. Evaluation on a carefully curated benchmark for assessing the writing capabilities of LLMs shows Weaver models of all sizes outperform generalist LLMs several times larger than them. Notably, our most-capable Weaver Ultra model surpasses GPT-4, a state-of-the-art generalist LLM, on various writing scenarios, demonstrating the advantage of training specialized LLMs for writing purposes. Moreover, Weaver natively supports retrieval-augmented generation (RAG) and function calling (tool usage). We present various use cases of these abilities for improving AI-assisted writing systems, including integration of external knowledge bases, tools, or APIs, and providing personalized writing assistance. Furthermore, we discuss and summarize a guideline and best practices for pre-training and fine-tuning domain-specific LLMs.

Keywords

Cite

@article{arxiv.2401.17268,
  title  = {Weaver: Foundation Models for Creative Writing},
  author = {Tiannan Wang and Jiamin Chen and Qingrui Jia and Shuai Wang and Ruoyu Fang and Huilin Wang and Zhaowei Gao and Chunzhao Xie and Chuou Xu and Jihong Dai and Yibin Liu and Jialong Wu and Shengwei Ding and Long Li and Zhiwei Huang and Xinle Deng and Teng Yu and Gangan Ma and Han Xiao and Zixin Chen and Danjun Xiang and Yunxia Wang and Yuanyuan Zhu and Yi Xiao and Jing Wang and Yiru Wang and Siran Ding and Jiayang Huang and Jiayi Xu and Yilihamu Tayier and Zhenyu Hu and Yuan Gao and Chengfeng Zheng and Yueshu Ye and Yihang Li and Lei Wan and Xinyue Jiang and Yujie Wang and Siyu Cheng and Zhule Song and Xiangru Tang and Xiaohua Xu and Ningyu Zhang and Huajun Chen and Yuchen Eleanor Jiang and Wangchunshu Zhou},
  journal= {arXiv preprint arXiv:2401.17268},
  year   = {2024}
}
R2 v1 2026-06-28T14:32:13.454Z