PulseAugur
EN
LIVE 11:27:13

New CTR prediction methods boost performance via adaptive compute and gradient allocation

By PulseAugur Editorial · [4 sources] ·

Two new research papers propose novel methods for improving Click-Through Rate (CTR) prediction models. The first paper introduces UTTSI, a framework that dynamically scales inference compute based on instance uncertainty, leading to a 5.3% CTR gain in an A/B test. The second paper presents HeteGenCTR, which addresses gradient imbalance in generative CTR models by reallocating training weights to more difficult feature fields, showing significant improvements, especially for cold-start users. AI

IMPACT These novel approaches to CTR prediction could lead to more efficient and accurate ad targeting, improving user experience and advertiser ROI.

RANK_REASON Two academic papers published on arXiv proposing new methods for CTR prediction.

Read on arXiv cs.IR (Information Retrieval) →

AI-generated summary · Google Gemini · from 4 sources. How we write summaries →

COVERAGE [4]

  1. arXiv cs.AI TIER_1 English(EN) · Moyu Zhang, Yun Chen, Yujun Jin, Jinxin Hu, Yu Zhang, Xiaoyi Zeng ·

    Selective Test-Time Compute Scaling for Click-Through Rate Prediction via Uncertainty-Triggered Feature Path Exploration

    arXiv:2605.24989v1 Announce Type: cross Abstract: Scaling test-time compute has proven highly effective for language models, yet this opportunity remains largely unexplored for industrial Click-Through Rate (CTR) prediction. CTR models suffer from a fundamental asymmetry: feature…

  2. arXiv cs.LG TIER_1 English(EN) · Moyu Zhang, Yun Chen, Yujun Jin, Jinxin Hu, Yu Zhang, Xiaoyi Zeng ·

    Self-Balancing Gradient Allocation for Heterogeneity-Aware Feature Generation in Click-Through Rate Prediction

    arXiv:2605.24986v1 Announce Type: cross Abstract: Generative pre-training via discrete diffusion provides dense reconstruction supervision across all feature fields simultaneously, mitigating representation collapse from data sparsity in CTR prediction. However, all existing gene…

  3. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Xiaoyi Zeng ·

    Selective Test-Time Compute Scaling for Click-Through Rate Prediction via Uncertainty-Triggered Feature Path Exploration

    Scaling test-time compute has proven highly effective for language models, yet this opportunity remains largely unexplored for industrial Click-Through Rate (CTR) prediction. CTR models suffer from a fundamental asymmetry: feature combinations well-represented in training yield c…

  4. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Xiaoyi Zeng ·

    Self-Balancing Gradient Allocation for Heterogeneity-Aware Feature Generation in Click-Through Rate Prediction

    Generative pre-training via discrete diffusion provides dense reconstruction supervision across all feature fields simultaneously, mitigating representation collapse from data sparsity in CTR prediction. However, all existing generative CTR methods share a fundamental limitation:…

RELATED TOPICS