Causal World Models Overview
• 1 min read 1 min
What I worked on
Looked into how LLMs and world models can reason causally about decisions. Compared causal ML, GFlowNets, and energy-based models.
What I noticed
- Bengio’s work links causality and representation learning
- GFlowNets connect causal reasoning and generative modeling
- JEPA and energy-based models share some conceptual overlap
”Aha” Moment
na
What still feels messy
Need more clarity on how causal inference actually interacts with representation learning inside these models.
Next step
Read Bengio’s causal world model papers and compare their learning objectives to GFlowNets.