Helix improves recommendation ranking by scaling feature and sequence models
HELIX: Purified and Unified - Rethinking Feature Interaction and Sequence Modeling for Large-Scale Recommendation
Information Retrieval
Summary
Recommendation systems try to guess what you want by looking at your past behavior and other information. The authors found that improving just one part of this guesswork, like looking only at features or only at past actions, doesn’t work very well. They designed a method called HELIX that combines both parts smoothly, making the system better at predicting what users want. When used in TikTok’s shopping recommendations, it helped show better videos and increased sales.
What this means in practice
- •For e-commerce platform engineers: Improve online product recommendations by jointly enhancing feature and user behavior sequence modeling for better ranking accuracy and sales.$Commercial implications: Allows e-commerce sites to increase user engagement and sales through more precise personalized recommendations.
- •For video streaming service developers: Enhance video recommendation systems by integrating user interaction features with long-term behavior sequences for improved viewer engagement.
Authors
Yuntao Zheng, Miao Zhang, Yadong Ding, Yanchuan Tang, Lixiyu Chen, Hao Wang, Quan Li, Shiying Cai, Yue Lin, Jiayu Li, Yu Feng, Wentao Yang, Rongkun Xing, Jiekai Wang, Mingge Zhang, Feiling Gong, Xiang Gao, Jinyu Dong, Yajing Zhang, Pengfei Ren, Yinzhou Wang
Abstract
Industrial recommendation ranking models typically scale along two modeling axes: feature interaction over heterogeneous user, item, context, and cross features, and sequence modeling over long, informative, and multi-type user behavior histories. We find that scaling either capability in isolation is insufficient, as each exhibits a limited scaling ceiling and a suboptimal scaling-law slope. We conjecture that achieving a more favorable scaling-law slope requires jointly scaling both axes. To support this, we present HELIX, a purified and unified architecture for large-scale recommendation. HELIX interleaves sequence retrieval and feature interaction while enforcing one-way information flow from reusable sequence states to candidate-conditioned mix-tokens. This design preserves cross-depth communication between the two modeling axes while keeping user-side sequence computation amortizable, enabling flexible and asymmetric scaling of sequence modeling and feature interaction. Deployed in TikTok's e-commerce recommendation system, HELIX consistently improves offline CTR AUC, CVR AUC, and other ranking metrics. In online A/B tests, it achieves an approximately 6% increase in e-commerce video GMV per user.