Yujian Betterest Li (bebetterest[at]outlook.com)
【x】 【github】 【scholar】 【wechat】
I am currently a master student (2023-2026), supervised by Kai Wu, in Artificial Intelligence Department at Xidian university in China, where I received my Bachelor's degree in June 2023.
Welcome to take a trip in my fascinating hometown.
BTW, I'm also a severe alcopilotic (a person addicted to copilot) 🤯
INFJ-T (Introverted – 100%, Intuitive – 93%, Feeling – 78%, Judging – 65%, Turbulent – 67%) for now (2024.09.23).
I am obviously a selfish and bigoted introvert, but I am trying to be a helpful and kind person most of the time... Hope that it would be better🧐
I do not have a specific field for research or a specific goal in my life yet, but I would like to try anything I am interested in and curious about, which may just be the meaning of my life. I have done some things — a small portion somewhat good, but most rather poor... For example, over the past year (2024), struggling with all my might for the opportunity to explore and try with my ideas, I still could not find the chance at all (though I got something else).
NOW: I'm looking for a potential position where I could share my minds and do some exciting things with other fascinating teammates!
Feel free to contact via email or Calendly. (I'm not good at speaking in crowds, but I'm absolutely talkative online or in one-on-one conversations🤖)
Hope that all of us, including those who dislike me and those I dislike, can live happily ever after.
Publications
Rapid Plug-in Defenders. (NeurIPS 2024, poster). Kai Wu, Yujian Betterest Li, Jian Lou, Xiaoyu Zhang, Handing Wang, Jing Liu. arxiv (Except for writing and revising the paper, almost everything was done by myself)
Pretrained Optimization Model for Zero-Shot Black Box Optimization. (NeurIPS 2024, poster). Xiaobin Li, Kai Wu, Yujian Betterest Li, Xiaoyu Zhang, Handing Wang, Jing Liu. arxiv (I just helped on the design of the network)
Some Other Experiences
(2025.02.24) I'm really proud that Wenxuan could do such an interesting thing under my little advice. Wenxuan did almost all of this by himself. 👍 arxiv
(2024.11-2025.02) Remote internship at MSRA on reasoning(math/code). Lucky to meet xinyu, yifei, jiahang, li... I was asked on many things, including mcts, self-correction, long synthetic cot, online rl... Due to constant requests to change direction, it is broad but superficial💀 I do not think that I deserve the 'Stars of Tomorrow' certificate😓
My very bitter lesson: I am not GPT; I do not like to simply produce inefficient outputs in a hurry based on discrete, fragmented, and arbitrary instructions and requirements; I need time and space for inference scaling to start from one point to coherently try, explore, expand, and iterate.
(2024.03-2024.07) Internship at Tencent AILab on LLM application for game. Happy to meet c, wang, tensor, peixin, ziv... We contributed to an llm-powered chat npc in a game and it's online now🕹️
(2023.11) JJQA: a Chinese QA dataset on the lyrics of JJ Lin's songs (GitHub Repository)
(2023.10) I tried to implement evolution algorithm, considering a large language model as a text generator, for global and coherent prompt optimization. (SPELL: Semantic Prompt Evolution based on a LLM)
(2023.08) In a competition on click-through rate prediction, with the given data, I tried to apply cross-attention and something similar to moe😅 to enhance feature fusion. (EMOFM: Ensemble MLP mOdel with Feature-based Mixers for Click-Through Rate Prediction) (9th)
(2022.05) In a competition on performance estimation, we tried to apply a mini transformer encoder for feature extraction and a multitask gaussian process model for regression. This is my first time collaborating and communicating with someone I met online🤗. (GitHub Repository) (9th)
"...每一颗心都有最向往的地方...当未来在那黑夜划出一道光..."
---《最向往的地方》
