AAAI submission in preparation
Embodied Failure-Recovery Data Collection
A data-collection and evaluation workflow for recoverable embodied task failures in AI2-THOR, with replayable traces and branch-level evidence.
Jiachang Zhang · AI Research
World models
for agents in
the real world.
Undergraduate researcher at XJTU and Research Intern at SJTU DMCV Lab. I work on embodied AI, world models, and data for learning systems.
Selected work
Experiments with visible evidence.
Ongoing research
LingBot Video MoE Routing
Controlled analysis of routing behavior and targeted gate interventions in a video mixture-of-experts model.
Research map
Three connected questions.
Embodied failure recovery
Collecting auditable trajectories that preserve failures, recovery attempts, and counterfactual branches.
Video world models
Studying how routed video models represent visual changes, dynamics, and physical interactions.
Multi-agent data
Contributing interaction data and quality control for world-model research in multi-agent settings.