Shresth Grover

prof_pic.jpg

Visiting Researcher, PRISM Lab

Princeton University

shgrover@ucsd.edu | +1 (619) 953-7137

My research focuses on developing multimodal foundation models capable of complex temporal reasoning to build intelligent, general-purpose agents. I am interested in designing novel architectures and training strategies that enable models to understand, plan, and act in dynamic environments, with a particular emphasis on enhancing robustness, generalization, and corrective behavior.

selected publications

  1. EgoLAP: Learning from Egocentric Human Data through Language-Action Reasoning
    L. Zha, Shresth Grover, T. Yin, and 7 more authors
    arXiv 2026
    arXiv
  2. OverLay++: Dense-Overlap Layout-to-Image Generation Dataset
    S. Aggarwal, Shresth Grover, D. Srivastava, and 7 more authors
    NeurIPS 2026
    arXiv
  3. CosPlan: A Benchmark for Corrective Sequential Planning
    Shresth Grover, Vibhav Vineet, and Yogesh S. Rawat
    ECCV 2026
    arXiv
  4. Enhancing Generalization in VLA Models by Preserving Pretrained Representations
    Shresth Grover, A Gopalkrishnan, Bo Ai, and 3 more authors
    IROS 2026
    arXiv Website
  5. How Do Inpainting Artifacts Propagate to Language?
    P. Yashwante, D. Abrahamyan, Shresth Grover, and 1 more author
    ACL 2026
    arXiv
  6. CoDiffusion: Harmonizing Layouts and Image Generation through Joint Diffusion
    Shresth Grover, S. Aggarwal, D. Srivastava, and 1 more author
    Under Review
  7. Navigating Hallucinations for Reasoning of Unintentional Activities
    Shresth Grover, Vibhav Vineet, and Yogesh S. Rawat
    EMNLP 2024
    arXiv
  8. Revealing the Unseen: Benchmarking Video Action Recognition under Occlusion
    Shresth Grover, Vibhav Vineet, and Yogesh S. Rawat
    NeurIPS 2023 2023
    PDF