Tencent UNI-REC Challenge Solution Boosts Ad pCVR Prediction

This paper presents the solution developed for the KDD Cup 2026 Tencent UNIREC Challenge, which focused on jointly modeling multi-domain user behavior sequences and non-sequential multi-field features to predict target-ad pCVR (predicted Conversion Rate). The proposed model, named Field-Aware RankMixer (FA-RankMixer), incorporates a dual-stream bilinear fusion mechanism. The FA-RankMixer begins by utilizing target-aware Deep Interest Network (DIN) modules to extract user interests from various behavior domains. It also distinguishes between recent and earlier interests for the longest behavior sequences. Semantic tokens are then formed based on feature fields and behavior domains, which are processed by RankMixer blocks for cross-token interaction. The architecture includes two complementary streams: a deep RankMixer stream and a shallow MLP stream. Their representations are then fused using a group-wise bilinear module. This comprehensive approach led to the solution ranking ninth on the official leaderboard for the challenge. The code for this solution has been made publicly available.