ALBERT WU
> studying cs + math @ UW-Madison
research
A compilation of the research I have done on Machine Learning, Data-Centric AI, and Formally Verifiable Agents.
A. Wu, N. Roberts, H. Lin, T.-H. Huang, S. Cho, and F. Sala, “A Multiagent Framework for Safe LLM Coding with Formal Guarantee”, to be submited to NeurIPS Workshop: AI for Verifiable Coding, 2026.
Extension of the ProD SAFe project to establish safety guarantees across terminal and CUDA settings.
N. Roberts, A. Wu, T.-H. Huang, and F. Sala, “ProD SAFe: Programmatic Distillation for Safe and Assured Foundation Models & Robots”, as part of a federal contract, Sep 2025 - May 2026.
A framework for safe and effective foundation model–powered robotics through model distillation, formal verification, and weak supervision.
N. Roberts, S. Cho, Z. Gao, T.-H. Huang, A. Wu, G. Orlanski, A. Trost, K. Buchanan, A. Albarghouthi, and F. Sala, “Test-Time Scaling Makes Overtraining Compute-Optimal”, in Conference On Language Modeling (COLM), 2026.
A new Train-to-Test (T^2) scaling laws that jointly optimize both pre-training and deployment time decisions to train the best model under any end-to-end budget constraint.
A. Wu, B. Hill, and J. Hanna, “Small LLM RLVR Optimization”, Fall 2025.
Graduate RL class final project - Investigating whether RLVR post-training can be stabalized and applied on LLMs with less than 7B parameters.