DACT

This is the PyTorch implementation of our paper: Drift-Aware Continual Tokenization for Generative Recommendation

Overview

We propose DACT, a Drift-Aware Continual Tokenization framework with two stages:

1. Tokenizer Fine-tuning: Augmented with a jointly trained Collaborative Drift Identification Module (CDIM) that outputs item-level drift confidence and enables differentiated optimization for drifting and stationary items.
2. Hierarchical Code Reassignment: Using a relaxed-to-strict strategy to update token sequences while limiting unnecessary changes.

DACT Tokenizer

Train

cd rqvae
bash finetune_book_dact.sh

Tokenize

cd rqvae
bash generate_code_dact.sh

Instantiation

For TIGER

cd TIGER-backbone
bash finetune.sh

For LC-Rec

cd LC-Rec-backbone/scripts_qwen
bash train.sh