Masked-feature reconstruction with discrete diffusion, followed by CTR fine-tuning, improves click prediction over discriminative baselines in offline and online tests.
In Proceedings of the 1st Workshop on Deep Learning for Recommender Systems (DLRS@RecSys)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DGenCTR: Towards a Universal Generative Paradigm for Click-Through Rate Prediction via Discrete Diffusion
Masked-feature reconstruction with discrete diffusion, followed by CTR fine-tuning, improves click prediction over discriminative baselines in offline and online tests.