Vision–Language–Action
Learning generalist robot policies that connect visual understanding, language instructions, and closed-loop action.
Robotics MSc · National University of Singapore
Building robust intelligence for robots that see, predict, and act.
I’m Jiacheng Zhang, a researcher working at the intersection of embodied AI, vision–language–action models, latent world models, and robot learning.
4.44 / 5.0NUS MSc GPA
Top 3Undergraduate cohort
Dean’s ListAcademic distinction
Research system
One agenda, three interacting layers.
Hover to hold the three-dimensional orbit.
Latent World Models
Building predictive latent representations for long-horizon planning, fast replanning, and model-based control.
Robust Robot Learning
Improving reliability under camera shifts, sensor noise, visual perturbations, and unseen task conditions.
Selected work
From latent world models to reliable manipulation.
My work asks how robots can preserve high-level intent while adapting their actions to new viewpoints, perturbations, and dynamic environments.
01
Multi-view VLA · JD Logistics
CrossWAM
A dual-view extension of LaWAM that aligns third-person and wrist-camera observations to improve manipulation robustness under visual distribution shifts.
98.5%LIBERO success
77.4%LIBERO-Plus
02
World-action models · JD Logistics
PSI-WAM
A slow–fast VLA framework with persistent stage-level manipulation intent and rapid action replanning across changing conditions.
80.2%Seven perturbation families
83.2%RoboTwin 2.0
03
Multi-agent latent world models · NUS MARMot Lab
Role-Aware MA-TD-MPC
Centralized latent dynamics, role-conditioned representations, cross-agent attention, and residual consensus for coordinated tugboat–barge control.
+13.0%Final evaluation return
04
LLM agents · Singapore–MIT SMART
RoboAtlas
An agent-based research pipeline for multi-source robotics landscape mapping, structured extraction, automated verification, and human review.
ICRA ’26Late Breaking Result · Accepted
Publications & manuscripts
Research in motion.
Work spanning robotics landscape intelligence, world-action models, and cooperative model-based reinforcement learning.
01
The RoboAtlas: Mapping the Global Robotics Landscape
2026
02
PSI-WAM: Persistent Stage-Aware Intent from World Action Models for Robot Manipulation
03
SMaRT-Tug: Structured Multi-Agent Reinforcement Learning for Physics-Based Tugboat–Barge Collaborative Manipulation
2026
About
Research grounded in systems that have to work.
I am pursuing an MSc in Robotics at the National University of Singapore, following a First Class BSc (Hons) in Computer Science (AI) from the University of Liverpool, where I graduated in the top three of my cohort.
Across academic labs and industry teams, I have built end-to-end training and evaluation pipelines, developed model-based control algorithms, and tested robot learning systems under systematic distribution shifts.
A+ Research ProjectNUS MSc Robotics
First ClassBSc (Hons), CS & AI
SingaporeOpen to PhD opportunities