Xidian University
B.Eng. · OMEGA Lab
Rank 4 / 174 · National Scholarship
Yue Su
Experience
A continuous path through
research and robotics.
MVIG · SJTU
Research Assistant
Robot learning
Astribot
Research Intern
Whole-body manipulation
ByteDance Seed
Research Intern
Seed Robotics
MMLab · HKU
PhD
World models & action
Selected works
Four projects, one research thread.
01MBA
Motion Before Action
Object motion becomes the manipulation condition.
02Dense Policy
Dense Policy
Bidirectional autoregressive learning of actions.
03DSPv2
Whole-body policy
Effective perception, generalization, and coherent action.
04World Guidance
World Guidance
World modeling in condition space for action generation.
MBA · 02 / 02
Two diffusion processes, one manipulation condition.
Step 1
Motion diffusion
Predict object motion from the current observation and language instruction.
Step 2
Action diffusion
Condition the manipulation policy on the generated motion representation.
Property
Plug and play
Attach MBA to existing manipulation policies without redesigning the action backbone.
Dense Policy · 01 / 02
Action generation is a design choice.
Sparse first.
Dense later.
Dense Policy predicts a few structural keyframes, then expands them bidirectionally in logarithmic inference steps.
Dense Policy · 02 / 02
Expand from sparse keyframes—in both directions.
Bidirectional generation
Sparse → dense.
Predict key actions first, then recursively fill the intervals to form a coherent trajectory.
ICCV 2025 paper ↗Real-robot behavior
Coherent manipulation.
The policy preserves global task structure while refining detailed local actions.
See all demos ↗DSPv2 · 01 / 02
Whole-body mobile manipulation changes the problem.
Perception
See what matters
Mobile settings introduce larger visual variation and longer interaction horizons.
Coordination
Move as one system
Base, torso, and arm actions must form a coherent whole-body trajectory.
Generalization
Leave the training setup
A useful policy must transfer across objects, tasks, and environments.
DSPv2 · 02 / 02
Perceive globally. Act coherently.
World Guidance · One view
World modeling inside the condition space.
Compact
Less is more
Compress future observations into a non-redundant predictive space.
Predictable
Action-centric latent
The VLA predicts conditions that are intrinsically relevant to action.
Fine-grained
Guidance that acts
Future-aware conditions support precise and generalizable manipulation.
World Guidance · Method
Condition in. Condition out.
Complete inference.
Stage I
Condition in
Inject future observations into the action inference pipeline.
Stage II
Condition out
Predict compressed future conditions alongside actions.
Transfer
Guidance retained
Carry future knowledge into the VLA without requiring future frames at test time.
World Guidance · Human videos
Any video can provide a condition.
1920hhuman manipulation
video
video
220h11.5% annotated
with actions
with actions