Yewen Pu
Assistant Professor at NTU specializing in communicative and trustworthy AI systems
Summary
Work
Education
Writing
mrCAD: Multimodal Refinement of Computer-aided Designs
January 1, 2025Work on multimodal refinement of CAD models, indicating research in the intersection of AI and design, likely improving computer-aided design processes.
Hypothesis search: Inductive reasoning with language models
January 1, 2023Explores using language models for inductive reasoning through hypothesis searching.
Communicating Natural Programs to Humans and Machines
January 1, 2022A publication exploring methods for AI systems to communicate with humans and machines in a natural way, focusing on interactive systems and task alignment through dialogue.
Fusion 360 gallery: A dataset and environment for programmatic cad construction from human design sequences
January 1, 2021Introduces a dataset and environment for programmatic CAD construction, highlighting work on converting human design sequences into CAD models.
Program Synthesis with Pragmatic Communication
January 1, 2020Research on program synthesis incorporating pragmatic communication, enabling AI systems to generate programs more effectively by considering human intent and feedback.
Write, execute, assess: Program synthesis with a repl
January 1, 2019Details a system for program synthesis that involves writing, executing, and assessing programs within a Read-Eval-Print Loop (REPL) environment.
Making Fast Informative Queries with Learned Propagations
January 1, 2019Yewen Pu's PhD thesis, which explores how to generate quick and informative queries using learned propagation techniques.
Verifiable reinforcement learning via policy extraction
January 1, 2018Research focusing on making reinforcement learning processes more transparent and trustworthy through the extraction of verifiable policies.
Inversecsg: Automatic conversion of 3d models to csg trees
January 1, 2018Presents a method for automatically converting 3D models into Constructive Solid Geometry (CSG) trees.