There's a growing worry that AI will break empirical social science -- that agents can p-hack until they find something that "works." We think that worry deserves to be taken seriously. Our new paper shows that is true empirically and makes it precise: njw.fish/static/paper...

New paper: "Distribution-Conditioned Transport" Modern scientific datasets don't contain one population — they contain thousands. Clones, donors, patients, each its own distribution. DCT learns transport maps that generalize across them, including to distributions never seen during training.

This project was accepted at #NeurIPS2025! We will be presenting today at 4:30 in Exhibit Hall C,D,E Poster Location: #1805! Swing by and learn about multiscale modeling, distribution representation learning, and some cool statistics!

🚨 New preprint 🚨 We introduce Generative Distribution Embeddings (GDEs) — a framework for learning representations of distributions, not just datapoints. GDEs enable multiscale modeling and come with elegant statistical theory and some miraculous geometric results! 🧵

Excited to share our new work on causal sensitivity analysis for fairness metrics at #NeurIPS2024! We've developed a causal sensitivity analysis framework to understand how underlying measurement biases (encoded by DAGs) impact machine learning fairness evaluations. 1 / 5

I’m at NeurIPS! If you’re in Vancouver and want to catch up let me know! I’ve been working at the intersection of generative models and causal inference, focused on bio/healthcare. If anything from active learning to causal foundation models sound interesting I’d love to chat!

i guess its time to start posting here more regularly?

another day writing a custom qp solver, another day doing more linear algebra than could possibly be healthy

are there any active set methods for least squares that allow for equality and inequality constraints while preserving sparsity? i have not found any...