Research map / Architectures and efficiency / Quantization
Adaptive Low-Precision Training for Embeddings in Click-Through Rate Prediction
Shiwei Li, Huifeng Guo, Lu Hou, Wei Zhang, Xing Tang, Ruiming Tang, Rui Zhang, Ruixuan Li
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| learn a feature-wise step size instead of using a fixed step size (ablation) | LPT(SR) to form ALPT(SR) · Avazu | AUC: 0.7927 -> 0.7951 | Improved |
| learn a feature-wise step size instead of using a fixed step size (ablation) | LPT(SR) to form ALPT(SR) · Criteo | AUC: 0.8123 -> 0.8144 | Improved |
| use stochastic rather than deterministic rounding (ablation) | ALPT · Criteo | AUC: 0.8096 -> 0.8144 | Improved |
| use stochastic rather than deterministic rounding (ablation) | LPT · Avazu | AUC: 0.7654 -> 0.7927 | Improved |
| use stochastic rather than deterministic rounding (ablation) | LPT · Criteo | AUC: 0.7966 -> 0.8123 | Improved |
| use feature-wise adaptive step sizes rather than fixed step sizes (ablation) | LPT(SR) · Avazu 2-bit | AUC: 0.7829 -> 0.7851 | Improved |
| use feature-wise adaptive step sizes rather than fixed step sizes (ablation) | LPT(SR) · Criteo 2-bit | AUC: 0.8009 -> 0.8033 | Improved |
| increase embedding dimension to 32 (ablation) | ALPT(SR) · Avazu, d=32 | AUC: 0.7951 -> 0.7955 | Mixed |
| use ALPT instead of LPT (ablation) | CTR embedding quantization · Criteo, threshold=2 | AUC: 0.8115 -> 0.8126 | Improved |
| use 2-bit ALPT instead of 2-bit LSQ | CTR embedding quantization · Avazu 2-bit | AUC: 0.7912 -> 0.7851 | Hurt |
| use 2-bit ALPT instead of 2-bit LSQ | CTR embedding quantization · Criteo 2-bit | AUC: 0.8089 -> 0.8033 | Hurt |
Closest papers by meaning
- QuEST: Stable Training of LLMs with 1-Bit Weights and Activations (ICML 2025 · similarity 0.85)
- ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization (NeurIPS 2025 · similarity 0.84)
- QT-DoG: Quantization-Aware Training for Domain Generalization (ICML 2025 · similarity 0.84)
- Exploring Post-training Quantization in LLMs from Comprehensive Study to Low Rank Compensation (AAAI 2024 · similarity 0.84)
- AMPA: Adaptive Mixed Precision Allocation for Low-Bit Integer Training (ICML 2024 · similarity 0.84)
- HESTIA: A Hessian-Guided Differentiable Quantization-Aware Training Framework for Extremely Low-Bit LLMs (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
