Reproduce “Structured Learning of Compositional Sequential Interventions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jialin Yu, Andreas Koukorinis, Nicolò Colombo, Yuchen Zhu, Ricardo Silva
The CSI-VAE predicts outcomes under intervention sequences that a unit may not have experienced. It represents baseline behavior with latent features, applies action-specific changes that depend on time since each action, and composes those changes with recent behavior and fixed covariates. In Spotify-based semi-synthetic simulations, its reported error score is 68.23 versus 253.85 for an ablation at T+1; interventions were artificial.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.