Profile

About me

I am a master's student in Software Engineering at Peking University. My research spans LLM post-training, reinforcement learning for agents, LLM evaluation and reward modeling, diffusion language models, and multimodal or embodied intelligence. I firmly believe that research should start with real business needs and return to serve them. I have two first-author papers and one co-authored paper accepted at conferences rated Category A by Tsinghua University.

My research aims to move language models from generating answers to taking reliable action: enabling autoregressive and diffusion models to learn from verifiable feedback, improve through tools, vision, and real-world interaction, and translate capability gains into measurable value.

At Tencent CSIG, WorkBuddy Team × Youtu Lab, I research Code Agent post-training, with a focus on evaluation improvements for WorkBuddy Bench, web-agent evaluation, training data, and reinforcement learning, connecting real user needs, capability diagnosis, and model training to enable continuous evolution through an iterative feedback loop.

Available Jul 2027 LLM Research Scientist · Research Engineer
Mainland China · Hong Kong · Singapore
Curriculum Vitae Google Scholar Citations GitHub Stars
01

Research directions

From failure modes to training, evaluation, and systems.

01 Model layer

Foundation Models & Post-training

Training objectives and inference methods that improve model capability without hiding failure modes behind aggregate scores.

SFTPreference OptimizationDiffusion LMs
02 Action layer

Agents & Reinforcement Learning

Credit assignment, tool use, and reward design for agents that must interact with environments rather than only generate text.

Tool UseGRPOCredit Assignment
03 Trust layer

Evaluation & Reward Modeling

Robust LLM judges, long-form evaluation, uncertainty, and reward models whose scores reflect what users actually care about.

LLM-as-a-JudgePRMLong-form
04 World layer

Multimodal, VLM & Embodied AI

Visual feedback and robot reward models that connect language-model reasoning with perception and action.

VLMVisual FeedbackRobot Rewards
02

Latest news

Research, publication, and career updates.

  1. Our paper Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning was accepted to NeurIPS 2026, and its paper, code, and model artifacts are now publicly available.

  2. Our paper Length-Adaptive Decoding for Masked Diffusion Machine Translation was accepted to the EMNLP 2026 Main Conference. The paper, code, data, and model weights are now publicly available.

  3. I am currently working on Code Agent post-training at Tencent CSIG, WorkBuddy Team × Youtu Lab, with a focus on evaluation improvements for WorkBuddy Bench, web-agent evaluation, training data, and reinforcement learning, connecting real user needs, capability diagnosis, and model training to enable continuous evolution through an iterative feedback loop.

  4. Our paper RefineSVG was accepted to ACM Multimedia 2026, a CCF-A conference.

  5. We released new work on tool-calling reinforcement learning, diffusion-language-model reasoning, and robot reward modeling.

  6. I joined Tencent PCG’s QQ Algorithm Group to work on multilingual large-language-model training and reinforcement learning for agents.

  7. I started my master’s study at Peking University.

03

Publications & preprints

“Research starts with real business needs.”

Showing all 6 papers.

EMNLP '26 CCF-B Accepted

During my advisor's collaboration with BYD

Length-Adaptive Decoding for Masked Diffusion Machine Translation

Yan Zhan, Mengkai Hou, Wanting Zhang, Zhijun Gao

Entropy-Valley is a training-free method for selecting a masked diffusion model's target canvas before denoising, improving translation adequacy without a learned length predictor.

EMNLP 2026 Main Conference

ACM MM '26 CCF-A Accepted

RefineSVG: Visual Feedback-Driven Reinforcement Learning for Image-to-SVG Generation

Shaobo Liu, Feiqiao Mao, Shuaishuai Zhou, Weiqi Tan, Zhiqiong Lu, Yan Zhan, Zhengping Liang

A closed-loop visual-feedback framework that lets multimodal language models render, compare, and iteratively correct SVG output, supported by an SVG-oriented vocabulary and agentic reinforcement learning.

ACM Multimedia (ACM MM) 2026

Under Review

Completed during my internship with the QQ Algorithm Group

SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL

Yan Zhan, Shaobo Liu, Qiunan Liu, Yuanjun Shi, Siqi Xu, WeiYi Hou, Xiang Xu, Zekang Li, Weizhou Pan, Jiahong Yan

Segment-Locked Credit Assignment routes execution rewards to tool tokens and preference rewards to summary tokens, reducing cross-segment credit contamination in tool-calling reinforcement learning.

Under Review

Under Review

With Beijing Innovation Center of Humanoid Robotics Co., Ltd.

TrustRoboReward: Preference-Ordered Isotonic Score Editing for Multi-Paradigm Robot Reward Models

Yidong Wang, Yan Zhan, Ziteng Feng, Zhenyu Cui, Ziyi Zhou, Renzhao Liang, Jiaxuan Zhu, Zilei Yang, Yiran Zhao, Zhongkuan Mao, Bo Jia, Hanchu Ni, Chenggang Xie, Biao Liu, Yi Zhang, Yong Dai, Xiaozhu Ju, Wei Ye, Shikun Zhang

A multi-paradigm robot reward-modeling framework whose POISE module calibrates pointwise scores to respect more reliable pairwise preferences.

Under Review

Under Review

During my advisor's collaboration with BYD

Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection

Yan Zhan, Yunze Song, Mengkai Hou, Wanting Zhang, Shaobo Liu, Zhijun Gao

Byte Authority studies how the same decoded bytes can reach an LLM agent as either one reserved control token or ordinary subword tokens, exposing how learned token representations influence prompt-injection authority.

Under Review

Under-review manuscripts are shown with their target venues; CCF labels describe venue class only. Classification follows the CCF seventh-edition directory. View directory

04

Industry experience

“Research returns to serve real business needs.”

Tencent CSIG · WorkBuddy Team × Youtu Lab

WorkBuddy Code Agent Algorithm Research Intern

Beijing, China · On-site

Lead evaluation-driven Code Agent post-training for WorkBuddy, connecting capability diagnosis, targeted training data, and reinforcement learning. Built separate functional and aesthetic evaluation, difficulty-based data selection, and quality control for outputs and teacher trajectories; integrated SGLang with browser sandboxes and developed probability-aware GRPO clipping for low-probability tool-call tokens. Improved expert agreement to 90.7% and reduced training time by approximately 74% in the reported comparison. Browser-based verification also accelerated reward computation, and the evaluation pipeline now supports model iteration and training decisions, connecting real user needs, capability diagnosis, and model training to enable continuous evolution through an iterative feedback loop.

Code Agentpost-trainingRLtool useWorkBuddy× Youtu Lab

Tencent PCG · QQ

Large Language Model Algorithm Research Intern

Shenzhen, China · On-site

Built a one-million-pair multilingual translation corpus for QQ International spanning 13 Eurasian languages, covering data filtering and cleaning, low-resource language-pair distillation, NER entity repair, 4B-class base-model evaluation, and SFT. The resulting model improved in-domain performance by approximately 10 percentage points while matching a substantially larger MoE translation baseline on FLORES-200 and WMT25.

1Mtraining pairs13languages+10ppin-domain

Perfect World

NLP Algorithm Intern

Beijing, China · On-site

Developed an LLM-based advertising-slogan recognition system for player-community moderation. Combined retrieval-augmented generation with LoRA fine-tuning and deployed efficient inference with vLLM. The system reached an F1 score of 85% and supported hourly offline batch recognition.

85%F1 scoreRAG+ LoRAvLLMdeployment
05

Education

Degrees, majors, and academic performance.

Peking University

Expected

Master's in Software Engineering

GPA: 3.71/4.0. Coursework includes machine learning, natural language processing, large language model development, AI practice, and algorithm design.

GPA 3.71 / 4.0Expected Jun 2027

Beijing Forestry University

Bachelor's in Digital Media Technology

Overall rank: 2/37, including first place in the 2022–2023 academic year.

Overall rank 2 / 372022–2023 rank 1 / 37
06

Honors & awards

Competitions, scholarships, and student leadership.

First Prize, Beijing Regional Round, Higher Education Main Track

8th China International College Students' 'Internet+' Innovation and Entrepreneurship Competition

Second and Third Prizes, National Finals

National College Student Digital Media Technology Works and Creativity Competition

Second Prize

2021 National Modeling Competition

Academic Excellence Scholarship

Beijing Forestry University

Outstanding Student Scholarship

Beijing Forestry University

Arts and Sports Excellence Scholarship

Beijing Forestry University

Merit Student

Beijing Forestry University

Outstanding Student Organization Officer

Beijing Forestry University