Hi, I'm Jusheng Zhang
Homepage · Google Scholar · Email
My earlier GitHub accounts are no longer accessible. I’m gradually migrating and reorganizing my projects under this account.
My research connects reliable agents, efficient visual intelligence, and embodied decision-making. My goal is to help build the next generation of AI agents for scientific discovery.
- Stanford University — Visiting Researcher, working on agentic AI for life sciences.
- Google DeepMind — Student Researcher, working on efficient video and vision-language models.
- Sun Yat-sen University — Former Lab Researcher at HCP Lab, focused on multi-agent and multimodal learning.
- Agents — collaboration, uncertainty-aware reasoning, and learning from failures.
- Vision-language models — token compression, visual reasoning, and video representations.
- Robotics — cooperative planning and world-action modeling.
- AI for science — agents that support scientific research and discovery.
- Self-Rewarded Multimodal Coherent Reasoning Across Diverse Visual Domains
NeurIPS 2026 · Spotlight - Failure-Driven Workflow Refinement
ICML 2026 · Spotlight - Hybrid Token Compression for Vision-Language Models
CVPR 2026 · Highlight · Code
More code: GAM-Agent · CF-VLM · 3D-Agent
All publications · Google Scholar
- Nanyang Technological University — Research Intern.
- Meta AI — Research Intern, Llama 4 team.
- Snap Research — Research Collaborator.
- HCP Lab, Sun Yat-sen University — Lab Researcher.
- Ant Group — Risk Control Algorithm Expert, 2018–2023.
I am grateful to Prof. Keze Wang for giving me the opportunity to begin my research journey.
- Lead Organizer, LoViF Challenge on Efficient VLMs, CVPR 2026.
- Organizer, LoViF ActPhysCause Challenge, ECCV 2026.
- Reviewer for NeurIPS, ICML, ICLR, CVPR, ICCV, and ECCV, among others.
Email · Homepage · Google Scholar · LinkedIn

