MicroVerse Micro-World Simulation

Mechanism-aware world modeling for organs, cells, and subcellular biological processes.

Micro-world models
MicroVerse World models Biological simulation Mechanistic reasoning ICLR 2026
MicroVerse overview

MicroVerse explores a different kind of world model: not city streets or game worlds, but biological micro-worlds where organ, cellular, and subcellular processes unfold under domain mechanisms that are often invisible to ordinary vision-language models.

Research Storyline

Scope
Move world models into biology

MicroVerse treats organs, cells, and subcellular processes as evolving worlds whose important state is often hidden from surface images.

Mechanism
Use scientific constraints, not only pixels

The project asks whether generative systems can respect biological mechanisms and state transitions, not merely produce plausible-looking pictures.

Benchmark
Evaluate micro-world dynamics

Benchmarks pressure models to predict, generate, and reason about biologically meaningful transitions over time.

Bridge
Connect scientific and economic simulations

MicroVerse complements TwinMarket and Economic World Models by showing how the same simulation agenda can apply to high-stakes scientific domains.

Research Questions

Can AI simulate hidden biological dynamics?

Micro-world modeling requires reasoning about mechanisms, state transitions, and biological constraints, not only visual appearance.

How should micro-worlds be evaluated?

The project builds tasks and benchmarks that test whether generated or predicted dynamics remain biologically meaningful.

What does a scientific world model need?

The system points toward simulation models that can be inspected, corrected, and aligned with scientific evidence.

MicroVerse benchmark figure

Display Figures

Paper Trail

Micro
MicroVerse: A Preliminary Exploration Toward a Micro-World Simulation

Introduces biological micro-world simulation as a world-modeling direction for organs, cells, and subcellular processes.

Repository
Macro
Economic World Models and TwinMarket

Shows the same agentic simulation idea at social and market scale, giving MicroVerse a broader world-modeling context.

Economic World Models

Why It Matters

  • It expands world modeling from visually obvious environments to scientific domains where mechanisms are hidden and high-stakes.
  • It connects medical AI, biological modeling, and agentic simulation under one research question: can models reason over evolving worlds?
  • It complements Economic World Models and TwinMarket by showing that the same simulation agenda can apply to microscopic systems.