Thomas Walker

PhD Candidate in 3D Vision and Neural Graphics
University of Edinburgh & Generative AI Laboratory (GAIL)
Google Scholar | GitHub | LinkedIn | CV

About Me

I am a final-year Ph.D. candidate at the University of Edinburgh. My current research focus is improving the physical realism of interactive world models. In particular, persistent memory, viewpoint consistency, causality, and interactivity. I have a formal background in mathematics and first-author publications at top-tier CV/ML conferences.

Selected Publications

• PERSIST: World Models with Persistent 3D State
  Thomas Walker*, Samuel Garcin*, Steven McDonagh, Tim Pearce, Hakan Bilen, Kaixin Wang, Tianyu He, Jiang Bian. ICML, 2026. [Website]
  *Equal Contribution
• 
  Semantic Foam: Unifying Spatial and Semantic Scene Decomposition
  Amr Sharafeldin, Shrisudhan Govindarajan, Thomas Walker, Aryan Mikaeili, Daniel Rebain, Kwang Moo Yi, Andrea Tagliasacchi. CVPR, 2026. ⭐ Highlight ⭐ [Website]
• DynaSlice: Multi-Class 4D Reconstruction from Cross Sections
  Thomas Walker, João Paulo Silva do Monte Lima, Diana Aldana Moreno, Oisín Nolan, Salvatore Esposito, Tiago Novello. 2026. [Preprint]
• 
  CrossSDF: Reconstruction Of Thin Structures From Cross Sections
  Thomas Walker*, Salvatore Esposito*, Daniel Rebain, Amir Vaxman, Arno Onken, Changjian Li, Oisin Mac Aodha. CVPR, 2025. [Arxiv]
  *Equal Contribution
• 
  Spatially Adaptive Hash Encodings For Neural Surface Reconstruction
  Thomas Walker, Octave Mariotti, Amir Vaxman, Hakan Bilen. WACV, 2025. [Arxiv]
• Explicit Neural Surfaces: Learning Continuous Geometry With Deformation Fields
  Thomas Walker, Octave Mariotti, Amir Vaxman, Hakan Bilen. NeurIPS NeurReps Workshop, 2023. [Arxiv]
• 
  Spherical Feature Pyramid Networks
  Thomas Walker, Varun Anand, Pavlos Andreadis. arXiv, 2023. [Arxiv]