Rethinking Generative Recommender Tokenizer: Recsys-Native Encoding and Semantic Quantization Beyond LLMs
从信息论角度剖析 SID 的设计理念
Blogs
A Rank-One Popularity Component in Dot-Product Recommender Scores
推荐 Softmax 目标中的 Rank-One Popularity Component
R3-VAE: Reference Vector-Guided Rating Residual Quantization VAE for Generative Recommendation
更稳定的生成式推荐 Tokenizer: 正交残差分量 & Cosine & 聚类正则
HiD-VAE: Interpretable Generative Recommendation via Hierarchical and Disentangled Semantic IDs
用层次化解耦 VAE 生成可解释的语义 ID
Markovian Pre-Trained Transformer for Next-Item Recommendation
用合成马尔科夫链预训练推荐模型
BEAR: Beam-Search-Aware Optimization for LLM-based Recommendation
Beam Search 的训推一致约束学习
CoFiRec: Coarse-to-Fine Tokenization for Generative Recommendation
从粗到细的层级化语义 Tokenization
Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation
推荐系统中合成 (diverse) 数据对于 Scale 的重要性
Differentiable Semantic ID for Generative Recommendation
直接受推荐损失监督的 Tokenizer
Factorizing personalized Markov chains for next-basket recommendation
FPMC