Linear-PAL: A Lightweight Ranker for Mitigating Shortcut Learning in Personalized, High-Bias Tabular Ranking

TL;DR

Linear-PAL is a lightweight, structurally constrained learning framework that effectively mitigates shortcut learning in high-bias ranking scenarios, outperforming deep ensembles in de-biasing and efficiency.

Contribution

The paper introduces Linear-PAL, a novel de-biasing method with explicit feature conjunctions and regularization, plus a vectorized hashing technique for efficient feature generation.

Findings

Linear-PAL outperforms deep ensembles in de-biased ranking quality.

Linear-PAL reduces training latency by 43 times.

Linear-PAL enables near real-time, personalized ranking updates.

Abstract

In e-commerce ranking, implicit user feedback is systematically confounded by Position Bias -- the strong propensity of users to interact with top-ranked items regardless of relevance. While Deep Learning architectures (e.g., Two-Tower Networks) are the standard solution for de-biasing, we demonstrate that in High-Bias Regimes, state-of-the-art Deep Ensembles suffer from Shortcut Learning: they minimize training loss by overfitting to the rank signal, leading to degraded ranking quality despite high prediction accuracy. We propose Linear Position-bias Aware Learning (Linear-PAL), a lightweight framework that enforces de-biasing through structural constraints: explicit feature conjunctions and aggressive regularization. We further introduce a Vectorized Integer Hashing technique for feature generation, replacing string-based operations with $O(N)$ vectorized arithmetic. Evaluating on a large-scale dataset (4.2M samples), Linear-PAL achieves Pareto Dominance: it outperforms Deep Ensembles in de-biased ranking quality (Relevance AUC: 0.7626 vs. 0.6736) while reducing training latency by 43x (40s vs 1762s). This computational efficiency enables high-frequency retraining, allowing the system to capture user-specific emerging market trends and deliver robust, personalized ranking in near real-time.