αlphabell.
α · People · Wen Shao
Stylised avatar of Wen Shao

Wen Shao

Long-tenured contributor

Based in Shanghai
Node-cell hadamard-08
ORCID 0000-0157-6046-3438

Research

Wen leads hadamard-08, the cell working on cross-modal latent unification — the practical question of whether vision, language, and symbolic representations can share latent dynamics structure to a degree that produces transfer at scale. The cell is one of the smaller world-models cells and one of the more theory-leaning.

She is a co-author on the lab's compositional latent dynamics paper and has been with alphabell since 2020. Her work bridges the world-models axis and the interpretability axis; the cross-modal unification result has implications for what circuits compose in foundation models.

Wen is based in Shanghai. She is on the world-models cross-cell review pool and is one of the contributors most often paired with interpretability cells for unification-related work.

Background

Ph.D. computer science, Shanghai Jiao Tong University, 2014.

Prior to alphabell: SJTU; Volterra Cognition; Praxis AI Studies.

Selected publications

Full publications index →

Recent talks

  • Cross-modal latent unification at scale, NeurIPS 2025
  • What perception and prediction share, ICLR 2025
Working with

Wen is currently part of node-cell hadamard-08, working under the World models research axis. The cell is open to substantive correspondence from researchers working on adjacent problems; route requests through hadamard-08@alphabell.com or directly to Wen at wen-shao@alphabell.com.