Yifan Liu 刘一帆
Hi! I'm a PhD student at the University of Illinois Urbana-Champaign, supervised by Prof. Dong Wang. Previously, I earned an M.S. in Computer Science from Georgia Institute of Technology and a BSc in Computer Science (Hons) from University College London. I am always happy to discuss my research and explore potential collaborations, so please feel free to reach out if our interests overlap.
-
Long-Horizon Agent Safety
Safety drift and emerging risks in stateful, complex environments (e.g., computer-use agents).
-
Trustworthy ML
Fairness, interpretability, uncertainty quantification, and performance trade-offs.
-
LLM Post-Training
Diverse post-training paradigms for capability enhancement, adaptation, and alignment.
I am actively seeking U.S.-based research internship opportunities for Summer 2027 in areas related to agentic AI, AI safety, and trustworthy information retrieval.
News
- A paper on uncertainty quantification for LLM MAS has been accepted to EMNLP (Main) 2026! 🎉
- A paper on generative recommendation and item tokenization has been accepted to ACM RecSys 2026! 🎉
- Started a research internship at IBM Research. 👁️🐝𝐌
- A paper on cross-module interference in agentic systems has been accepted to the Failure Modes in Agentic AI Workshop at ICML! 🎉
- Two papers on generative recommendation and session-based recommendation were accepted to SIGIR 2026! 🎉
- Two papers on uncertainty estimation and image-generation red-teaming were accepted to WWW 2026!
- A journal paper on reasoning-based uncertainty estimation was accepted to Knowledge-Based Systems! 🎉
- A paper on designing explanations for misinformation detection was accepted to CSCW 2025! 🎉
- A paper on unauthorized individual image-generation protection was accepted to ICCV 2025! 🎉
- A paper on face-component fairness was accepted to ACM FAccT 2025! 🎉
- A paper on multimodal interactive MoE for fake news detection was accepted to WWW 2025! See you in Sydney! 🎉
- Released a new dataset for multidimensional media bias.
- A paper on retrieval-augmented fact verification was accepted to Findings of ACL 2024!
- Started my PhD at UIUC!
Work Experience
-
IBM ResearchMay 2026 – Aug 2026Research InternYorktown Heights, NY, USADeveloping methods to attribute and explain actions in LLM-based agent systems, with a focus on safety and accountability.
Publications
Query-Conditioned Agent Action Attribution with White-Box Proxies
Mitigating the Fairness Steering Tax via Adaptive Inference-time Subspace Intervention
PropUQ-MAS: Propagation-Aware Uncertainty Quantification for LLM Multi-Agent Systems
Red-Teaming Privacy-Protective Perturbations: Blind Face Restoration as an Attack Strategy
SPRINT: Scalable and Predictive Intent Refinement for LLM-Enhanced Session-based Recommendation
Retrieval Augmented Fact Verification by Synthesizing Contrastive Arguments
No publications match this filter.
Education
-
University of Illinois Urbana-ChampaignSep. 2023 – PresentPh.D. in Information SciencesAdvisor: Prof. Dong Wang -
Georgia Institute of TechnologySep. 2021 – May 2023M.S. in Computer ScienceAdvisor: Prof. Harish Ravichandar -
University College LondonSep. 2017 – May 2019BSc in Computer ScienceAdvisor: Prof. Mark Herbster
Teaching and Service
- IS 202: Social Aspects of Information Technology — Facilitated discussion sections and graded coursework.
- IS 305: Programming for Information Problems — Graded programming assignments and provided feedback.
- IS 597: Human-Centered Data Science — Designed assignments and delivered concept tutorials.
My Cats 🐾
