Yikai Zhao's Homepage

About me

I am an undergraduate student at Sun Yat-sen University (SYSU), pursuing a B.Sc. in Geographic Information Science and a B.Eng. in Computer Science and Technology. I expect to graduate in June 2027.

My research focuses on large language models, multimodal reasoning, and agents, with a particular interest in reliable and efficient multimodal intelligence.

I am currently conducting research on multimodal reasoning at the XAI Lab, where I am fortunate to be supervised by Prof. Weijiang Yu. I am also working on LLM reasoning and embodied harness at QLab, University of California San Diego, where I am fortunate to be supervised by Prof. Lianhui Qin.

Previously, I conducted research on computer vision and AI for geoscience at the IMARS Lab, Sun Yat-sen University, where I was fortunate to be supervised by Prof. Da He.

For research discussions and potential collaborations, feel free to contact me at zhaoyk25@mail2.sysu.edu.cn.

Research Interests

  • LLM Reasoning: Post-Training (e.g., On-Policy Distillation and Reinforcement Learning) and Inference-Time Decoding
  • Multimodal Reasoning
  • Agentic Systems and Self-Evolving Agents

News

Publications

† Equal Contribution   * Corresponding Author

2026

Overview of Context-Aware Cluster Decoding
Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs
Yikai Zhao, Qiyan Zhao, Jiaquan Zhang, Xiaofeng Zhang, Xiaosong Yuan, Pengzhou Cheng*
EMNLP 2026 Main Conference · August 2026
[Paper] [Code]
A training-free, block-free decoding method that uses semantic-anchor-driven context-aware cluster scoring to improve long-form coherence and reduce hallucination in diffusion multimodal large language models.
Remote Sensing paper
Investigating Very-High-Resolution Land Cover Mapping in the Pearl River Delta with Remote Sensing Foundation Models and Multi-Source Data Bayesian Fusion
Junshen Luo†, Yikai Zhao†, Mingyang Xuan, Jizhou Zheng, Yan Zhou*, Xiaoping Liu
Remote Sensing (SCI, Q2), 18(6), 897 · March 2026
[Paper] [Code] [Dataset]
My first co-first-authored paper! Congratulations to all co-authors! · Developed SDMAE foundation model and SBFNet; constructed PRD VHR land cover datasets; designed decision-level Bayesian fusion for complex delta mapping.
IEEE TIP paper
Robust Fine-Grained Oriented Ship Detection for Remote Sensing Imagery via Controllable Generative Pretraining
Da He, Haoran Liu, Zeyu Li, Yikai Zhao, Xikun Hu, Ping Zhong, Weijia Li, Qian Shi, Xiaoping Liu*, Yanfei Zhong, Liangpei Zhang
IEEE Transactions on Image Processing (TIP, CCF-A, SCI Q1 TOP) · June 2026
[Paper]
Congratulations to all co-authors! · Built LAFI large fine-grained ship dataset; proposed COSD generative pre-training framework with diffusion-based data augmentation for rotated ship detection.

2025

PSODNet paper
PSODNet: Pre-trained Scene-aware Object Detection for Optical Remote Sensing Imagery
Haoran Liu, Da He*, Yikai Zhao, Qian Shi, Xiaoping Liu
IEEE Transactions on Geoscience and Remote Sensing (TGRS, CCF-B, SCI Q1 TOP) · October 2025
[Paper] [Code]
My first paper! Congratulations to all co-authors! · Proposed PSODNet with multi-head pre-training EON network and SORM module to suppress complex background noise for multi-scale remote sensing object detection.

Honors & Awards

  • 🏅 Sun Yat-sen University Outstanding Undergraduate Scholarship (2024, 2025, 2026)
  • 🥈 2nd Prize, 15th Asia and Pacific Mathematical Contest in Modeling (APMCM) (2025)
  • 🥈 2nd Prize (Guangdong Provincial Division), 16th Lanqiao Cup National Software Competition — C/C++ Undergraduate Group A (2026)
  • 🥉 3rd Prize (Guangdong Provincial Division), 16th Chinese Mathematics Competitions (CMC) (2024)

Beyond Research

Outside of research, I enjoy playing basketball and football, and hitting the gym.

Visitor Map