Research map / Label-efficient and robust learning / Recognition and long-tail
Unified Embedding: Battle-Tested Feature Representations for Web-Scale ML Systems
Benjamin Coleman, Wang-Cheng Kang, Matthew Fahrbach, Ruoxi Wang, Lichan Hong, Ed H. Chi, Derek Zhiyuan Cheng
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add feature multiplexing (one shared table for all features) (ablation) | Hashing trick · Criteo 2.5MB | AUC: 79.44 -> 80.49 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | Hashing trick · Movielens-1M 158kB | AUC: 77.07 -> 82.00 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | Hashing trick · Avazu 324kB | AUC: 75.10 -> 76.86 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | QR embedding · Criteo 2.5MB | AUC: 79.74 -> 80.48 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | PQ embedding · Criteo 2.5MB | AUC: 79.82 -> 80.54 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | PQ embedding · Avazu 324kB | AUC: 76.48 -> 76.96 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | PQ embedding · Movielens-1M 158kB | AUC: 87.32 -> 88.09 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | ROBE-Z embedding · Avazu 324kB | AUC: 76.88 -> 76.94 | No clear effect |
| add feature multiplexing (one shared table for all features) (ablation) | Hash embedding · Movielens-1M 1.6MB | AUC: 88.28 -> 88.79 | Improved |
| add feature multiplexing (one shared table for all features) (ablation) | Hash embedding · Criteo 2.5MB | AUC: 80.40 -> 80.31 | Hurt |
| add feature multiplexing (one shared table for all features) (ablation) | QR embedding · Criteo 25MB | AUC: 80.29 -> 80.22 | Hurt |
| add feature multiplexing (one shared table for all features) (ablation) | Hashing trick · Criteo training speed (25MB, CPU) | steps/second: 29.2 -> 29.4 | No clear effect |
Closest papers by meaning
- Towards Universal Image Embeddings: A Large-Scale Dataset and Challenge for Generic Image Representations (ICCV 2023 · similarity 0.88)
- Large-Scale Heterogeneous Feature Embedding (AAAI 2019 · similarity 0.88)
- NUDGE: Lightweight Non-Parametric Fine-Tuning of Embeddings for Retrieval (ICLR 2025 · similarity 0.87)
- U-MARVEL: Unveiling Key Factors for Universal Multimodal Retrieval via Embedding Learning with MLLMs (ICLR 2026 · similarity 0.86)
- NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models (ICLR 2025 · similarity 0.86)
- When is an Embedding Model More Promising than Another? (NeurIPS 2024 · similarity 0.86)
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.
