Contact

✉️ zwl23 [at] mails.tsinghua.edu.cn

✉️ acha131441373 [at] gmail.com

📍 Beijing, China

Github: https://github.com/Achazwl

Google Scholar: ‪Weilin Zhao‬ - ‪Google Scholar‬

Twitter: Weilin Zhao (@acha_William_) / X

Hobbies

🏸 Badminton

📷 Photography

⛰️ Travel

Languages

🇨🇳 Chinese

🇨🇳 Cantonese

🇬🇧 English

<aside> <img src="/icons/brightness-high_blue.svg" alt="/icons/brightness-high_blue.svg" width="40px" /> I’m Weilin Zhao, a third-year PhD student in the THUNLP, Department of Computer Science and Technology, Tsinghua University.

</aside>

<aside> <img src="/icons/search_blue.svg" alt="/icons/search_blue.svg" width="40px" /> Research Interests: Efficient LLM

</aside>

Selected Publications

See more on my Google Scholar

• [ICLR 2026] InfLLM-V2: Dense-Sparse Switchable Attention for Seamless Short-to-Long Adaptation Weilin Zhao, Zihan Zhou, Zhou Su, Chaojun Xiao, Yuxuan Li, Yanghao Li, Yudi Zhang, Weilun Zhao, Zhen Li, Yuxiang Huang, Ao Sun, Xu Han, Zhiyuan Liu PDF, CODE
• [ACL 2025] FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling Weilin Zhao, Tengyu Pan*, Xu Han, Yudi Zhang, Ao Sun, Yuxiang Huang, Kaihuo Zhang, Weilun Zhao, Yuxuan Li, Jianyong Wang, Zhiyuan Liu, Maosong Sun* PDF, CODE
• [EMNLP 2024] Ouroboros: Generating Longer Drafts Phrase by Phrase for Faster Speculative Decoding Weilin Zhao, Yuxiang Huang*, Xu Han, Wang Xu, Chaojun Xiao, Xinrong Zhang, Yewei Fang, Kaihuo Zhang, Zhiyuan Liu, Maosong Sun* PDF, CODE
• [COLM 2024] CA-LoRA: Adapting Existing LoRA for Compressed LLMs to Enable Efficient Multi-Tasking on Personal Devices Weilin Zhao, Yuxiang Huang*, Xu Han, Zhiyuan Liu, Zhengyan Zhang, Kuai Li, Chen Chen, Tao Yang, Maosong Sun.* PDF, CODE
• [COLM 2025] BlockFFN: Towards End-Side Acceleration-Friendly Mixtureof-Experts with Chunk-Level Activation Sparsity Chenyang Song∗, Weilin Zhao∗, Xu Han, Chaojun Xiao, Yingfa Chen, Yuxuan Li, Zhiyuan Liu, Maosong Sun PDF, CODE
• [SC 2025] BurstEngine: an Efficient Distributed Framework for Training Transformers on Extremely Long Sequences of over 1M Tokens Ao Sun, Weilin Zhao*, Xu Han*, Cheng Yang, Zhiyuan Liu, Chuan Shi, Maosong Sun* PDF, CODE
• [NAACL 2025] Seq1F1B: Efficient Sequence-Level Pipeline Parallelism for Large Language Model Training Ao Sun, Weilin Zhao*, Xu Han, Cheng Yang, Zhiyuan Liu, Chuan Shi, Maosong Sun* PDF, CODE
• [ACL 2022 Best Demo] OpenPrompt: An open-source framework for prompt-learning Ning Ding, Shengding Hu*, Weilin Zhao*, Yulin Chen, Zhiyuan Liu, Haitao Zheng and Maosong Sun* PDF, CODE

Projects

• Maintainer: ‣ Lightweight Inference framework for Big Models

  • Implement a highly optimized attention kernel for InfLLM v2 in MiniCPM4
  • Implement Marlin GPTQ kernel
  • Implement the Speculative Sampling method in https://github.com/thunlp/FR-Spec

• MiniCPM Deployment

  • Deploy MiniCPM-V 1.0 on Android phones using MLC-LLM https://github.com/OpenBMB/mlc-MiniCPM

  • Deploy MiniCPM-V 2.0 on Android phones and PC using Llama.cpp https://github.com/ggerganov/llama.cpp/pull/6919

  • Support MiniCPM 3 on server using SGLang https://github.com/sgl-project/sglang/pull/1371 https://x.com/zhyncs42/status/1833444567959408876

• Maintainer: https://github.com/OpenBMB/BMTrain Efficient Training Framework for Big Models

• Maintainer: https://github.com/thunlp/OpenPrompt An Open-Source Framework for Prompt-Learning.

Experience

Undergraduate Student at Department of Computer Science and Technology, Tsinghua University

2019 - 2023

PhD Student at THUNLP, Tsinghua University

2023 - Present

Under the guidance of Associated Professor Zhiyuan Liu. Mainly working with Yuxiang Huang and Xu Han.