Shicheng Fan

📧 Email / 🎓 Google Scholar / 😺 Github / 💼 LinkedIn

👋 I am a first-year CS Ph.D. student (Aug. 2025–) at University of Illinois Chicago (UIC), advised by Prof. Lu Cheng, collaborating with Prof. Kun Zhang (CMU). I received my B.Eng. from Zhejiang University (Chu Kochen Honors College), where I worked on dexterous robotic manipulation with Prof. Qi Ye.

My research interests center on Causality and its applications in AI for Science, Reinforcement Learning, and Large Language Models.

Shicheng Fan

🎉 News

  • [2026.01] New preprint: TRACE — Trajectory Recovery for Continuous Mechanism Evolution in Causal Representation Learning. 📄 Paper
  • [2025.08] Started Ph.D. at UIC! 🎓

🎓 Education

University of Illinois Chicago , Ph.D. in Computer Science , 08/2025 - present
Zhejiang University , B.Eng. in Automation (Control & Robotics) , 09/2021 - 06/2025
Chu Kochen Honors College · GPA: 3.98/4.0 · Rank: 6/121 (Top 5%) · Outstanding Graduate

📝 Publications and Preprints

First-author

Collaborative Publications

  • Hanyu Su, Shicheng Fan, Yifan Cui, Lu Cheng. Conformalized Proximal Causal Inference with a Single Proxy. Under Review

🔬 Research Experience

  • TRACE: CRL under Continuous Mechanism Evolution · First Author · 2025–2026
    • Proposed the first theoretical framework extending causal representation learning from discrete mechanism shifts to continuous transitions; proved joint identifiability of latent causal variables and continuous mixture trajectories via nonlinear ICA and variational inference
    • Designed the TRACE framework based on Mixture-of-Experts, where each expert learns an atomic mechanism and time-varying mixture coefficients recover mechanism trajectories at inference
  • Knowledge-Enhanced RL for LLM Factuality Alignment · Co-1st Author · 2025–present
    • Designing an end-to-end LLM post-training pipeline combining SFT and RL for improved factuality alignment
  • Causal Representation Learning for Protein Conformational Transitions · 1st Author · 2025–present
    • Applying causal representation learning to molecular dynamics for identifying key residues driving protein conformational transitions
  • Semantic and Goal-Driven Dexterous Hand Manipulation · B.Sc. Thesis · 2024–2025
    • Reproduced the Text2HOI three-stage framework, matching reported performance on H2O, GRAB, and ARCTIC datasets
    • Proposed a Shadow Hand → MANO cross-embodiment retargeting algorithm via skeleton-point alignment and two-stage gradient optimization
    • Joint training with GraspM3 dataset (8000+ objects) improved Physical Realism to 0.897 and reduced penetration volume by ~50%

🏆 Awards & Honors

  • Outstanding Graduate, Zhejiang University, 2025
  • Academic Excellence Model, Zhejiang University, 2022, 2023, 2024
  • Outstanding Student in AI&ML Summer Project, University of Cambridge, 2023
  • National Second Prize, Advanced Mathematics Competition of China, 2022

🏅 Competitions

  • ASC (Asia Supercomputer Community Challenge), National Second Prize, 2024
    • Built full HPC system from hardware to software; optimized LLM inference with KV Cache & batching strategies
  • ACM-IPCC (International Parallel Computing Challenge), National Second Prize, 2022
    • Optimized program runtime from 13min to 200ms (~4000x speedup) using MPI, CUDA, SIMD