Eohan G.

👋 Hi, I’m Eohan, a senior undergraduate student at Central South University, and I have been admitted to Shanghai Jiao Tong University for a direct Ph.D program. Prior to that, I had the privilege of being supervised by Prof. Xiyao Liu, where I conducted research on Deepfake Detection.

💭 My research interests: AI Agents (Current Focus), Multimodal Large Language Models(MLLMs), Embodied Intelligence, and Deepfake Detection (Previous Works).

📢 Seeking Collaboration:I am actively looking for like-minded partners for academic research collaboration. If interested, please feel free to hit me up via email at gnonymous.ai@gmail.com.

🔥 News

• [2024.12] Honored with National Scholarship for the 2023-2024 academic year (Top 1%). NEW
• [2024.12] Honored with Special Scholarship @ CSU for the 2023-2024 academic year (Top 1% & Only 1). NEW
• [2024.09] Having been admitted to SJTU for a direct Ph.D Program.NEW
• [2024.08] 🎉 Congratulations to our project “Deepshield: A Robust Defense Against Deepfake” for winning the National Second Prize(Top 6-10%) in the National College Student Information Security Competition(Totally 852 teams participated).
• [2024.05] Finish my research intern @ University of Michigan-SJTU Joint Institute: Exploring MLLMs in Human-Vehicle Dialogue.
• [2024.03] 🎉 Congratulations to our Paper “Multi-strategy Adversarial Learning for Robust Face Forgery Detection under Attacks” has accepted to ICME 2024 as Oral Paper (CCF-B Conference).
• [2024.03] 🎉 Congratulations to our project for receiving funding from College Student Innovation Project in Hunan Province, which was awarded “Outstanding Project & Paper” among 4 other accolades. I’m the leader of this project.
• [2022-2024] Honored with Excellent Student and Excellent Student Cadre of CSU(Totally 800+ students in the grades).
• [2022.12] Honored with National Scholarship for the 2021-2022 academic year (Top 1% & Only 1)

📝 Publications

ICME 2024 CCF-B Oral Paper

Multi-Strategy Adversarial Learning for Robust Face Forgery Detection Under Heterogeneous and Composite Attacks
Xiyao Liu¹, Fengkai Dong¹, Xin Liao¹, Yuhan Guo¹, Jianbiao He¹, Jian Zhang¹, Gerald Schaefer², Hui Fang².
1.Central South University 2.Loughborough University

🎖 Honors and Awards

• [2024.12] National Scholarship for 2023-2024 academic year(Top 1%).
• [2024.12] Special Scholarship for 2023-2024 academic year(Top 1% & Only 1), Central South University.
• [2024.08] National Second Prize (Top 6-10%), National College Student Information Security Competition.
• [2024.03] Outstanding Project & Paper Award(Leader of team), College Student Innovation Project in Hunan Province.
• [2023.12] First Class Scholarship for 2022-2023 academic year(Top 5%), Central South University.
• [2022-2024] Excellent Student * 2 and Excellent Student Cadre(Totally 800+ students in the grades), Central South University.
• [2022.12] National Scholarship for 2021-2022 academic year(Top 1% & Only 1).
• [2022.12] First Class Scholarship for 2021-2022 academic year(Top 5%), Central South University.

📖 Educations

• 2021.09 - now, Undergraduate, Central South Univeristy, Changsha.

💬 Blogs 11 articles

• [2025.02] “DeepScaleR: Surpassing O1-Preview with a 1.5B Model by Scaling RL”: The Path to Empowering Small Models with Inference Capabilities via RL. NEW
• [2025.02] “Paper Reading: UI-TARS: Pioneering Automated GUI Interaction with Native Agents”: An end-to-end GUI agent that achieves superior performance in automated GUI interactions through enhanced perception, unified action modeling, system 2 reasoning, and iterative training. NEW
• [2025.01] “Paper Reading: Offline Reinforcement Learning for LLM Multi-Step Reasoning”: Proposing OREO, an offline RL method that improves LLM reasoning by jointly optimizing policy and value functions through the soft Bellman equation. NEW
• [2024.04] “YOLO in Autonomous Driving: A Systematic Review”: Analyzing YOLO’s evolution and applications in autonomous vehicles.
• [2024.04] “Vision Transformer vs CNN”: Exploring their fundamental differences and unique characteristics in visual tasks
• [2024.04] “Vision Transformer”: Exploring their fundamental differences and unique characteristics in visual tasks.
• [2024.04] “Deepfake Detection Survey”: Covering detection methods and future challenges.
• [2024.03] “Lips Don’t Lie”: Exploring innovative approaches in face forgery detection using lip movement analysis.
• For more, please visit Blog’s Category.

👨‍💻 Internships

• [2024.05-2024.06] As a research intern @ University of Michigan-SJTU Joint Institute: Exploring MLLMs in Human-Vehicle Dialogue.