MATT-CTR: Unleashing a Model-Agnostic Test-Time Paradigm for CTR Prediction with Confidence-Guided Inference Paths

TL;DR

MATT introduces a model-agnostic test-time paradigm for CTR that mitigates low-confidence feature combinations by generating multiple instance-specific inference paths guided by confidence scores. Confidence is estimated via Hierarchical Probabilistic Hashing, which bounds frequencies of feature combinations and isolates high-frequency cases with a min-heap, while low-frequency cases are modeled probabilistically to derive a lower bound. Confidence-guided path generation then samples and aggregates predictions across multiple high-confidence paths, improving robustness without retraining. Empirical results on four datasets and an online A/B test show consistent gains over strong baselines and practical deployment viability, highlighting inference-time optimization as a valuable complement to training-time improvements in CTR systems.

Abstract

Recently, a growing body of research has focused on either optimizing CTR model architectures to better model feature interactions or refining training objectives to aid parameter learning, thereby achieving better predictive performance. However, previous efforts have primarily focused on the training phase, largely neglecting opportunities for optimization during the inference phase. Infrequently occurring feature combinations, in particular, can degrade prediction performance, leading to unreliable or low-confidence outputs. To unlock the predictive potential of trained CTR models, we propose a Model-Agnostic Test-Time paradigm (MATT), which leverages the confidence scores of feature combinations to guide the generation of multiple inference paths, thereby mitigating the influence of low-confidence features on the final prediction. Specifically, to quantify the confidence of feature combinations, we introduce a hierarchical probabilistic hashing method to estimate the occurrence frequencies of feature combinations at various orders, which serve as their corresponding confidence scores. Then, using the confidence scores as sampling probabilities, we generate multiple instance-specific inference paths through iterative sampling and subsequently aggregate the prediction scores from multiple paths to conduct robust predictions. Finally, extensive offline experiments and online A/B tests strongly validate the compatibility and effectiveness of MATT across existing CTR models.

Figures (4)

• Figure 1: An illustration of previous methods and ours. Features in dashed boxes represent candidate features available for selection. (a) Previous approach utilizes the full set of input features for prediction. (b) Our method iteratively constructs a high-confidence inference path via probabilistic sampling, using the selected feature combination for scoring.
• Figure 2: The structure of Model-Agnostic Test-Time framework (MATT) for CTR models. The left part shows the overall MATT scoring process, and the right part shows the specific inference path generation process.
• Figure 3: Prediction performance of three variants of MATT under three backbones on four datasets.
• Figure 4: Prediction performance of MATT in different parameters under three backbones on the industrial dataset.