LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913051302690816 |
|---|---|
| author | Tang, Jiakai Zhang, Runfeng Wang, Weiqiu Liu, Yifei Wang, Chuan Chen, Xu Yang, Yeqiu Wu, Jian Jiang, Yuning Zheng, Bo |
| author_facet | Tang, Jiakai Zhang, Runfeng Wang, Weiqiu Liu, Yifei Wang, Chuan Chen, Xu Yang, Yeqiu Wu, Jian Jiang, Yuning Zheng, Bo |
| contents | Scaling Transformer-based click-through rate (CTR) models by stacking more parameters brings growing computational and storage overhead, creating a widening gap between scaling ambitions and the stringent industrial deployment constraints. We propose LoopCTR, which introduces a loop scaling paradigm that increases training-time computation through recursive reuse of shared model layers, decoupling computation from parameter growth. LoopCTR adopts a sandwich architecture enhanced with Hyper-Connected Residuals and Mixture-of-Experts, and employs process supervision at every loop depth to encode multi-loop benefits into the shared parameters. This enables a train-multi-loop, infer-zero-loop strategy where a single forward pass without any loop already outperforms all baselines. Experiments on three public benchmarks and one industrial dataset demonstrate state-of-the-art performance. Oracle analysis further reveals 0.02--0.04 AUC of untapped headroom, with models trained with fewer loops exhibiting higher oracle ceilings, pointing to a promising frontier for adaptive inference. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_19550 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction Tang, Jiakai Zhang, Runfeng Wang, Weiqiu Liu, Yifei Wang, Chuan Chen, Xu Yang, Yeqiu Wu, Jian Jiang, Yuning Zheng, Bo Information Retrieval Scaling Transformer-based click-through rate (CTR) models by stacking more parameters brings growing computational and storage overhead, creating a widening gap between scaling ambitions and the stringent industrial deployment constraints. We propose LoopCTR, which introduces a loop scaling paradigm that increases training-time computation through recursive reuse of shared model layers, decoupling computation from parameter growth. LoopCTR adopts a sandwich architecture enhanced with Hyper-Connected Residuals and Mixture-of-Experts, and employs process supervision at every loop depth to encode multi-loop benefits into the shared parameters. This enables a train-multi-loop, infer-zero-loop strategy where a single forward pass without any loop already outperforms all baselines. Experiments on three public benchmarks and one industrial dataset demonstrate state-of-the-art performance. Oracle analysis further reveals 0.02--0.04 AUC of untapped headroom, with models trained with fewer loops exhibiting higher oracle ceilings, pointing to a promising frontier for adaptive inference. |
| title | LoopCTR: Unlocking the Loop Scaling Power for Click-Through Rate Prediction |
| topic | Information Retrieval |
| url | https://arxiv.org/abs/2604.19550 |