I am a new master student at Carnegie Mellon University, Language Technologies Institue(MIIS). Before that, I recieved my bachlor degree at Shanghai Jiao Tong University, majoring in Software Engineering.
Research Interests
My research focuses on the intersection of LLMs and Agents. I am particularly interested in:
- Agent Systems(Code Agent, Multi-agent Systems, Memory, Tools, …)
- Multi-modal Foundation Models
- Trustworthy AI
I am actively seeking research opportunities on Agent systems and willing to be open to any research topics for directed study or RA!
Papers
* indicates equal contribution, † indicates corresponding author.
2025
A Survey of Self-Evolving Agents: On Path to Artificial Super Intelligence
Huan-ang Gao*, Jiayi Geng*, Wenyue Hua*, Mengkang Hu*, Xinzhe Juan*, Hongzhang Liu*, Shilong Liu*, Jiahao Qiu*, Xuan Qi*, Yifan Wu*, Hongru Wang*, Han Xiao*, Yuhang Zhou*, Shaokun Zhang*, Jiayi Zhang, Jinyu Xiang, Yixiong Fang, Qiwen Zhao, Dongrui Liu, Qihan Ren, Cheng Qian, Zhenhailong Wang, Minda Hu, Huazheng Wang, Qingyun Wu, Heng Ji, Mengdi Wang†
TL;DR: First comprehensive survey on self-evolving AI agents, establishing a roadmap toward Artificial Super Intelligence.
LastingBench: Defend Benchmarks Against Knowledge Leakage
Yixiong Fang*, Tianran Sun*, Yuling Shi, Min Wang, Xiaodong Gu† | EMNLP 2025 Findings
TL;DR: A framework to continuously reinforce and safeguard existing benchmarks against knowledge leakage in LLMs.
AttentionRAG: Attention-Guided Context Pruning in Retrieval-Augmented Generation
Yixiong Fang, Tianran Sun, Yuling Shi, Xiaodong Gu | ACL 2025 Workshop
TL;DR: Novel attention-based method to prune irrelevant context in RAG systems, improving efficiency without sacrificing accuracy.
Enhancing Speech Large Language Models for Deepfake Speech and Audio Detection through Feature Dropout and Dropin
Yupei Li*, Yixiong Fang*, Lucia Specia, Xiaodong Gu, Björn Schuller
TL;DR: Improved deepfake audio detection in speech LLMs using novel dropout and dropin techniques.
2024
From Uncertainty to Trust: Enhancing Reliability in Vision-Language Models with Uncertainty-Guided Dropout Decoding
Yixiong Fang*, Ziran Yang*, Zhaorun Chen, Zhuokai Zhao†, Jiawei Zhou† | Preprint
TL;DR: Novel decoding strategy using uncertainty-guided dropout to improve VLM reliability and reduce hallucinations.
CodeCipher: Learning to Obfuscate Source Code Against LLMs
Yalan Lin, Chengcheng Wan, Yixiong Fang, Xiaodong Gu | Preprint
TL;DR: Automated code obfuscation technique to protect source code from being understood by large language models.
Efficient computation of eigenvalues in diffusion maps: A multi-strategy approach
Yixiong Fang*, Weixi Yang | SPML 2024
TL;DR: Multi-strategy approach for efficient eigenvalue computation in diffusion maps, improving computational performance.
Work Experience
Emagen AI (Startup)
Head of North America | Current Position
TL;DR: Leading the development of Cagen, an AI-native user based organization platform.
Ant Group Co., Ltd. - OceanBase
SQL Intern | July 2024 - August 2024
TL;DR: Optimized utf8 parsing with SIMD for up to 20x performance increase, add southeast Asia region chatset, add user friendly external table feature.
Step AI
AI Research Intern | March 2024 - June 2024
TL;DR: Develop the early stage agent, including tools(like weathers) and RAG system; develop the openapi platform of Step AI.
China Merchants Securities Co., Ltd.
Quantitative Research Intern | January 2024 - February 2024
TL;DR: Built quantitative trading strategies and performed market analysis using machine learning techniques.
Changjiang Securities International
Quantitative Research Intern | July 2022 - August 2022
TL;DR: Developed algorithmic trading models and conducted financial data analysis for equity markets.
Teaching
- Teaching Assistant for Mthread GPU Open Course (2025 Spring)
- Teaching Assistant for AI1102: Intel’s Cutting-Edge AI Algorithms and Practices (2025 Spring)
- Teaching Assistant for SJTU Xuesen Challenge Winter Camp(AI) (2025 Winter)
Open Source Projects
- MathTranslate
- OceanBase
