Anbang Liu (Andrew)
Master of Computer Science
Northwestern University
Hi there! I am Anbang (Andrew) Liu, a Master’s student in Computer Science at Northwestern University, where I am very fortunate to be advised by Prof. Manling Li and mentored by Qineng Wang and Zihan Wang on research in reasoning and planning for multimodal intelligence. I am also grateful to be advised by Prof. Siwei Chen and Prof. Junhan Zhao at the University of Chicago, under the mentorship of Zechuan Zhang and Zongxin Yang, on research in AI for Science. In addition, I have had the privilege of collaborating with Prof. David Demeter and Prof. Han Liu at Northwestern University. I feel incredibly thankful for the guidance and support I have received from all of these fantastic mentors and collaborators! Before joining Northwestern, I earned my bachelor’s degree in Computer Science from the University of San Francisco. Please feel free to reach out via email!
Education
Research Interests
My research interests lie at the intersection of reasoning, planning, multimodal intelligence, and AI for Science.
- In reasoning and planning, I focus on multimodal foundation models, agentic reinforcement learning, spatial intelligence, world modeling, and embodied AI. I am especially interested in how language- and vision-based agents understand complex environments, make decisions, and interact with the physical world.
- In AI for Science, I am interested in computational biology, medical AI, and interpretability, with the goal of developing machine learning methods that support scientific discovery and help address real-world biomedical problems.
News
- Sep 2026. 📢 We released MindTopo, a benchmark for evaluating topological reasoning and planning in foundation models.
- Aug 2026. 🎉 Our paper FABRIC was accepted to MICCAI 2026 (PedAItrics Workshop) for an Oral Presentation!
- Aug 2026. 🎉 Our paper AI-Powered Medicinal Chemistry and Translational Drug Development was published in Chemical Society Reviews!
- Jul 2026. 📢 We released our survey paper, Progress Reward Modeling for Robotic Learning. It ranked #2 on Hugging Face Daily Papers.
- Jun 2026. 🎉 One paper was accepted to MMLS 2026!
- Jun 2026. 📢 We released ReCLIP, a transformer-based framework for modeling protein-protein interactions (PPIs) at residue resolution.
- Jun 2026. 🎉 AI in Medicinal Chemistry and Translational Drug Development was accepted to Chemical Society Reviews!
Research Experience
- Designed and implemented 7 Gym-compatible benchmark environments for embodied and image-based VLM evaluation, defining task protocols and reproducible pipelines to assess reasoning, robustness, and planning across diverse visual scenarios.
- Used the VAGEN framework to connect vision-based Gym environments with SFT/RL training pipelines, trained models with RL algorithms, and evaluated pre- and post-training accuracy changes across task settings.
- Designed a human annotation interface and coordinated the external annotation workflow, including guideline preparation, annotation quality review, and response analysis to validate benchmark quality and characterize model failure modes.
- Developed a sequence-based machine learning framework for biomolecular interaction modeling with downstream interpretability analysis for biological investigation.
- Conducted multimodal medical AI research on clinically relevant imaging and related data, developing and evaluating machine learning models for predictive tasks.
- Trained and deployed robotic control policies for manipulation tasks, collected teleoperated demonstrations for image-conditioned offline learning, and evaluated policy performance in simulation and real-world settings for task success, generalization, and sim-to-real transfer.
- Developed a data-centric vision pipeline for self-driving laboratories that integrated automatically collected real-world data with validated synthetic images for robust bubble detection in precision-sensitive pipetting.
Publications (* = equal contribution, † = co-corresponding author, § = project lead)
FABRIC: Privacy-Preserving Federated Distillation of Foundation Models for Multi-Institutional Pediatric Glioma Recurrence Prediction
Jiying Wang*, Anbang Liu*, Zheyi Ji, Philip Chikontwe, Jiancheng Yang, Biyue Zhu, Peter Pytel, Sudarshawn Damodharan, Ziyue Xu, Kun-Hsing Yu, Samuel L. Volchenboum, Junhan Zhao
Accepted to MICCAI 2026 (PedAItrics Workshop) · Oral Presentation.
AI-Powered Medicinal Chemistry and Translational Drug Development
Kaicheng U*, Sophia Meixuan Zhang*, Ziyu Yu*, Zechuan Zhang, Jianwei Zhang, Chang He, Anbang Liu, Rui Chen, Stella Wang, Lijie Yan, Shichao Ding, Lavonda Li, Zongxin Yang, Gao Xiao, Xushuai Zhang, Kaige Bao, Haohan Wang, Athanasios V. Vasilakos†, Junhan Zhao†, Siwei Chen†, Xingcai Zhang†
Chemical Society Reviews, 2026.
Teaching
Northwestern University
Evanston, IL
Peer Mentor for CS449(Deep Learning), CS461(Deep Learning for NLP)
Industry
Felloc
Seattle, WA
Software Development Engineer
University of San Francisco
San Francisco, CA
USFMobile Usability and Functional Tester
Web Development Engineer
Professional Services
Workshop Reviewer
NeurIPS 2026 Workshop WM PAI