🏗️ System DesignID:
pre-ranking-architecturePre-Ranking Lightweight Architecture
粗排轻量模型与向量相似度剪枝
🎯Core Definition
The Pre-Ranking (Match-Filtering) Architecture is the critical intermediate layer bridging multi-channel retrieval and heavy ranking; within a strict 5ms latency budget, it scores and filters candidates from multi-channel recall down to high-potential items for heavy ranking; standard architectures include: 1) Vector-based Pre-Ranking (dual-tower dot products); 2) Distilled Lightweight DNNs (e.g., Alibaba COLD: a compact 2-3 layer MLP sharing lower embedding tables with the heavy ranker); 3) Feature-Separated Cross Networks (FSCD).
💡Use Cases
High-concurrency recommendation and ad systems where multi-channel recall yields 50,000+ candidates exceeding heavy ranker compute budgets.
⚡Key Problems Solved
Sending 10,000+ items directly to heavy rankers triggers massive SLA timeouts and infrastructure cost explosions; naive truncation destroys multi-channel diversity; pre-ranking achieves 90%+ heavy-ranker overlap at only 5% compute cost.
🎯5 High-Frequency Exam Points
1
Compare Two-Tower Pre-ranking vs Lightweight Interactive Pre-ranking (COLD) across expressive power and compute latency?
2
When pre-ranker shares embedding tables with the heavy ranker, how to prevent pre-ranker gradients from corrupting embedding precision?
3
How does Knowledge Distillation train the lightweight pre-ranker using the heavy ranker's soft prediction scores as teacher targets?
4
Why does pre-ranking evaluation prioritize Top-K Hit Rate / Overlap with heavy ranking outputs alongside standalone AUC?
5
How do C++ serving engines leverage AVX-512 SIMD and INT8 quantization to maximize pre-ranking throughput over 1000 items in 5ms?
📖 In-depth Guide:📄 recommendation-system-design →
Updated 2026-08-14🎯
🚀 Start Card Practice ➔Test Your Knowledge: Practice Questions for "Pre-Ranking Lightweight Architecture"
Single choice pitfall questions with instant feedback and mistake tracking.