Paper accepted at NeurIPS 2026

Our paper, Learning Deployable Causal Action Geometry under Temporal Non-Stationarity, has been accepted at NeurIPS 2026.

The work studies continuous decision making from observational time series under temporal drift. It combines anchored link-scale contrasts, cross-fitted orthogonal pilots, and profiled morphology selection to learn action-response structure that remains useful for deployment.

I am the first author of this work. Thank you to my co-authors and collaborators!