Zikai Xiao

I am Zikai Xiao (肖子恺), a Ph.D. student in Computer Science and Technology at Zhejiang University (2022–2027), advised by Prof. Zuozhu Liu. I am currently a Research Intern at the Qwen Team, Alibaba, where I work on agentic reinforcement learning and world models with Fei Huang and Jianhong Tu.

My research centers on large language models, with a focus on:

LLM Agents & Environments — web-agent RL, world models, and generative environments for training general agents.
Long-Context LLMs — long-input understanding and verifiable long-form generation.

Recent work includes WebWorld (ICML 2026) and Qwen-AgentWorld on web and agentic world models; LongWeave (EMNLP 2025) and Positional Contrastive Decoding (ACL 2025) on long-context LLMs; and earlier work on federated long-tailed learning (Fed-GraB, NeurIPS 2023; FedLoGe, ICLR 2024).

Feel free to reach out at zikai@zju.edu.cn — see also my Google Scholar and GitHub.

selected publications

ICML

WebWorld: A Large-Scale World Model for Web Agent Training

Zikai Xiao, Jianhong Tu, Chuhang Zou, and 7 more authors

In Proceedings of the International Conference on Machine Learning (ICML), 2026

arXiv Bib Code Website

@inproceedings{xiao2026webworld,
  title = {WebWorld: A Large-Scale World Model for Web Agent Training},
  author = {Xiao, Zikai and Tu, Jianhong and Zou, Chuhang and Zuo, Yuxin and Li, Zhi and Wang, Peng and Yu, Bowen and Huang, Fei and Lin, Junyang and Liu, Zuozhu},
  booktitle = {Proceedings of the International Conference on Machine Learning (ICML)},
  year = {2026},
}

EMNLP

LongWeave: A Long-Form Generation Benchmark Bridging Real-World Relevance and Verifiability

Zikai Xiao, Fei Huang, Jianhong Tu, and 7 more authors

In Findings of the Association for Computational Linguistics: EMNLP, 2025

Bib HTML Code

@inproceedings{xiao2025longweave,
  title = {LongWeave: A Long-Form Generation Benchmark Bridging Real-World Relevance and Verifiability},
  author = {Xiao, Zikai and Huang, Fei and Tu, Jianhong and Wei, Jianhui and Ma, Wen and Zhou, Yuxuan and Wu, Jian and Yu, Bowen and Liu, Zuozhu and Lin, Junyang},
  booktitle = {Findings of the Association for Computational Linguistics: EMNLP},
  year = {2025},
}

ACL

Mitigating Posterior Salience Attenuation in Long-Context LLMs with Positional Contrastive Decoding

Zikai Xiao, Ziyang Wang, Wen Ma, and 5 more authors

In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL), 2025

Bib HTML Code

@inproceedings{xiao2025pcd,
  title = {Mitigating Posterior Salience Attenuation in Long-Context LLMs with Positional Contrastive Decoding},
  author = {Xiao, Zikai and Wang, Ziyang and Ma, Wen and Zhang, Yan and Shen, Wei and Yan, Wang and Gong, Luqi and Liu, Zuozhu},
  booktitle = {Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)},
  year = {2025},
}

ICLR

FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data

Zikai Xiao, Zihan Chen, Liyinglan Liu, and 6 more authors

In Proceedings of the International Conference on Learning Representations (ICLR), 2024

arXiv Bib Code

@inproceedings{xiao2024fedloge,
  title = {FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data},
  author = {Xiao, Zikai and Chen, Zihan and Liu, Liyinglan and Feng, Yang and Wu, Jian and Liu, Wanlu and Zhou, Joey Tianyi and Yang, Howard Hao and Liu, Zuozhu},
  booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)},
  year = {2024},
}

NeurIPS

Fed-GraB: Federated Long-tailed Learning with Self-Adjusting Gradient Balancer

Zikai Xiao, Zihan Chen, Songshang Liu, and 7 more authors

In Advances in Neural Information Processing Systems (NeurIPS), 2023

Bib HTML Code

@inproceedings{xiao2023fedgrab,
  title = {Fed-GraB: Federated Long-tailed Learning with Self-Adjusting Gradient Balancer},
  author = {Xiao, Zikai and Chen, Zihan and Liu, Songshang and Wang, Hualiang and Feng, Yang and Hao, Jin and Zhou, Joey Tianyi and Wu, Jian and Yang, Howard Hao and Liu, Zuozhu},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2023},
}