🔬Causal Models Need Causal Data - Xaira’s X-Cell model for Drug Discovery (Bo Wang & Ci Chu, Chief Discovery Officer & Chief AI Scientist)
Drug discovery AI hits scaling walls from data information limits, not model size or compute
“test loss flatlines after 1.5B parameters while training loss continues to drop as you scale, that tells you that your model is limited by the amount of information in your data”
Xaira's X-Cell team found that gene expression prediction models hit a hard scaling ceiling driven by data information density, not parameters or compute — the 3.1B model fell off the scaling trend entirely. Expanding training data by ~30x broke through the wall, suggesting that in specialized scientific domains, data curation matters more than model scaling. This is a concrete demonstration of the 'causal data for causal models' thesis and a signal for how AI-for-science will diverge from general LLM scaling playbooks.