Selected work

Representative Papers

Variance Reduction for Heavy-Tailed Monetization Metrics in Ranking Experiments via Post-Stratification

Jun 02, 2026

This study addresses the challenge of high variance and low statistical power in online ranking experiments, where revenue-related metrics often exhibit heavy-tailed distributions—particularly problematic under limited traffic conditions. The authors propose a novel integration of post-stratification with the CUPED (Controlled-experiment Using Pre-Experiment Data) method, leveraging pre-experiment covariates to jointly reduce variance for heavy-tailed reward metrics. This approach significantly enhances experimental sensitivity without requiring additional traffic. Empirical deployment at ShareChat demonstrates substantial variance reduction, yielding approximately a 45% decrease in the required sample size to achieve equivalent statistical confidence. The work also provides a systematic characterization of the method’s applicability conditions and practical implementation guidelines.

0 citationsRead paper

Dimension Mask Layer: Optimizing Embedding Efficiency for Scalable ID-based Models

Oct 17, 2025

To address the high memory footprint and deployment challenges of recommendation models caused by large-scale ID feature embeddings, this paper proposes a learnable Dimensional Masking Layer that dynamically prunes individual dimensions of embedding vectors, enabling fine-grained, adaptive compression of embedding dimensionality. Integrated into the Keras embedding lookup backend, the layer is trained end-to-end to jointly optimize masking parameters and model weights, thereby mitigating multicollinearity and suppressing overfitting. Extensive evaluations on multiple public benchmarks and online A/B tests demonstrate an average 40–50% reduction in embedding dimensionality, yielding significant decreases in model memory consumption and inference latency, while preserving core metrics—including CTR and AUC—at stable levels. To our knowledge, this is the first work to introduce a differentiable, dimension-level selection mechanism for industrial-scale ID embedding compression, achieving a favorable trade-off between model lightweighting and performance robustness.

0 citationsRead paper

Recent publications

Latest Papers

Variance Reduction for Heavy-Tailed Monetization Metrics in Ranking Experiments via Post-Stratification

Jun 02, 2026

This study addresses the challenge of high variance and low statistical power in online ranking experiments, where revenue-related metrics often exhibit heavy-tailed distributions—particularly problematic under limited traffic conditions. The authors propose a novel integration of post-stratification with the CUPED (Controlled-experiment Using Pre-Experiment Data) method, leveraging pre-experiment covariates to jointly reduce variance for heavy-tailed reward metrics. This approach significantly enhances experimental sensitivity without requiring additional traffic. Empirical deployment at ShareChat demonstrates substantial variance reduction, yielding approximately a 45% decrease in the required sample size to achieve equivalent statistical confidence. The work also provides a systematic characterization of the method’s applicability conditions and practical implementation guidelines.

0 citationsRead paper

Dimension Mask Layer: Optimizing Embedding Efficiency for Scalable ID-based Models

Oct 17, 2025

To address the high memory footprint and deployment challenges of recommendation models caused by large-scale ID feature embeddings, this paper proposes a learnable Dimensional Masking Layer that dynamically prunes individual dimensions of embedding vectors, enabling fine-grained, adaptive compression of embedding dimensionality. Integrated into the Keras embedding lookup backend, the layer is trained end-to-end to jointly optimize masking parameters and model weights, thereby mitigating multicollinearity and suppressing overfitting. Extensive evaluations on multiple public benchmarks and online A/B tests demonstrate an average 40–50% reduction in embedding dimensionality, yielding significant decreases in model memory consumption and inference latency, while preserving core metrics—including CTR and AUC—at stable levels. To our knowledge, this is the first work to introduce a differentiable, dimension-level selection mechanism for industrial-scale ID embedding compression, achieving a favorable trade-off between model lightweighting and performance robustness.

0 citationsRead paper