PERSONALIZATION · 2026-10-30

Personalization — Paper Daily, 2026-10-30

Today's track has a single paper, from product ranking in Taobao's search by photo. It looks at a very common practice: ranking models place several feature-crossing networks side by side, train each on its own, and merge them at the end. The authors let these branches correct each other during training and constrain their hidden vectors to stay related without becoming copies, and online CTR and transactions both improve.

4.55
LEAD STORY · Alibaba

1. Multi-Branch Cooperation Networks for Enhanced Click-Through Rate Prediction in Large-Scale E-Commerce Search

Alibaba

Ranking teams should read this one, because it points out an often overlooked waste in multi-branch models: the branches are good at different samples, yet never learn from each other in training. The Pailitao team at Taobao (search by photo) makes three branches correct each other during training. On a given sample, if one branch gets the click right and another gets it wrong, the right branch's prediction is used to teach the wrong one. An extra constraint requires that one branch's hidden vector, multiplied by an orthogonal matrix, be close to the other branch's vector. In an online A/B test, CTR rose from 10.37% to 10.46% and GMV grew 1.62% relative. The gain from multiple branches comes not just from adding branches but from letting them fix each other where they disagree.

📖 Read the deep dive →