Unified Semantic and ID Representation Learning for Deep Recommenders
混合距离用于量化匹配 & 端到端的联合训练
2025
On the Markovian Nature of the Next-Item Recommendation Task
序列推荐任务的马尔科夫性
Generative Recommendation with Semantic IDs: A Practitioner’s Handbook
分析比较了现有生成式推荐的 Tricks 并给出了一个训练框架
Revisiting Self-attention for Cross-domain Sequential Recommendation
利用多任务/多目标优化学习到更佳的 Attention 分布, 促进跨域推荐
Bridging Textual-Collaborative Gap through Semantic Codes for Sequential Recommendation
通过 Q-Former 将 semantic IDs 转换为 textual IDs
TokenFlow: Unified Image Tokenizer for Multimodal Understanding and Generation
兼顾 Low-level 的 Pixel 信息和 High-level 的 Semantic 信息
UniCode$^2$: Cascaded Large-scale Codebooks for Unified Multimodal Understanding and Generation
非常自然的 Image-Codeword+Text-LLM-NextCodeWord-Generation 流程
Dynamic Chunking for End-to-End Hierarchical Sequence Modeling
符号序列的自动切分, 探究非 Subword Tokenizer 的可能性
SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizer
TiTok 框架 + Softmax 版 Vector Quantization 以期更高的压缩比
EARN: Efficient Inference Acceleration for LLM-based Generative Recommendation by Register Tokens
通过减小 KV cache size 加速 LLMRec 的推理