Solution Overview

This paper presents our solution to the KDD Cup 2026 Tencent UNI-REC Challenge. The task requires joint modeling of multi-domain user behavior sequences and non-sequential multi-field features for target ad click-through rate (pCVR) prediction.

Model Architecture

We developed a Field-Aware RankMixer (FA-RankMixer) with dual-stream bilinear fusion. The model first applies target-aware DIN modules to extract user interests from multiple behavior domains. It also separately models recent and earlier interests for the longest behavior sequence.

The model then forms semantic tokens based on feature fields and behavior domains, using RankMixer blocks for cross-token interaction. A shallow MLP stream complements the deep RankMixer stream, and a group-wise bilinear module fuses their representations.

Ultimately, our solution ranked ninth on the official leaderboard. Our code is available at GitHub.

Blogger's Review: The FA-RankMixer effectively combines multi-domain behavior features through its dual-stream architecture, demonstrating potential for precise modeling in complex data environments, warranting further exploration and optimization in practical applications.