TL;DR: HiRank is a recommender ranking model that fuses sequential and non-sequential features two ways in a single backbone, and uses effective-rank-aware attention to keep representations diverse. In an online test on the Taobao homepage, orders per user rose 2.27%.
Title: Unified Heterogeneous Feature Interaction via Effective-Rank Scaling for Recommender Systems
URL:
Key points
ใป๐งฉ Drops the encode-then-interact pipeline; event-level sequence states stay accessible across layers
ใป๐ช Models the main behavior sequence with HSTU-style gated attention
ใป๐ Retrieves from long-term history memories with both interest queries and candidate queries for diverse retrieval
ใป๐ Fuses sequence, retrieval outputs, features, and the candidate via token mixing
ใป๐ฏ Applies High-Rank Attention over scenario-aware compressed composite tokens
ใป๐ Offline: homepage GAUC 0.7801 (+0.0186 over the strongest baseline)
ใป๐ Online: orders/user +2.27%, GMV/user +2.86%
๐ก I like the takeaway that sequence length and representation diversity matter, not just model size.
#
RecSys# #
MachineLearning#