Publications
A full list is also available on Google Scholar.
arXiv preprint arXiv:2610.03516
A world action model that trains RGB, actions, depth, surface normals, and functional role segmentation as RGB videos with one video diffusion transformer, improving closed-loop success on held-out RLBench tasks.
Transactions on Machine Learning Research (TMLR)
A comprehensive survey examining the data infrastructure supporting Vision-Language-Action models in robotics.
- 2026 ROCKET: Residual-Oriented Multi-Layer Alignment for Spatially-Aware Vision-Language-Action Models
Advances in Neural Information Processing Systems (NeurIPS 2026)
A residual-oriented multi-layer alignment framework that injects 3D spatial reasoning into Vision-Language-Action models. Accepted to NeurIPS 2026.
ICML 2025 Workshop on Collaborative and Federated Agentic Workflows
A hierarchical memory framework that enables agents to leverage cross-domain experience for improved problem-solving capabilities.
arXiv preprint arXiv:2512.03381
Analyzing situated language use and ad-hoc convention formation within a large-scale corpus from a virtual 3D environment.
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL)
A student modeling approach for question generation to enhance test item analysis through simulation.