Generator Classifier

Generator-classifier models combine generative and discriminative approaches to improve various machine learning tasks. Current research focuses on enhancing efficiency and addressing limitations of autoregressive generators, particularly in applications like recommendation systems and language modeling, by exploring non-autoregressive alternatives and incorporating contrastive learning techniques. These models aim to improve generation quality while mitigating issues like toxicity and repetitiveness in text generation, and enhancing accuracy in zero-shot learning scenarios. The resulting improvements in efficiency and performance have significant implications for real-world applications, including personalized recommendations and improved natural language processing systems.

Papers

February 10, 2024

Non-autoregressive Generative Models for Reranking Recommendation
Yuxin Ren, Qiya Yang, Yichun Wu, Wei Xu, Yalong Wang, Zhiqiang Zhang
Effective Recommendation Modern Recommender System Two Stage Recommender Non Autoregressive Generative Generator Classifier

June 15, 2022

DIRECTOR: Generator-Classifiers For Supervised Language Modeling
Kushal Arora, Kurt Shuster, Sainbayar Sukhbaatar, Jason Weston
Language Model Intelligent Director Generator Classifier

April 24, 2022

Deconstructed Generation-Based Zero-Shot Model
Dubing Chen, Yuming Shen, Haofeng Zhang, Philip H. S. Torr
Structured Document Generalized Zero Shot Learning Zero Shot Model Unseen Test Distribution Generator Classifier

Generator Classifier

Papers

Non-autoregressive Generative Models for Reranking Recommendation

DIRECTOR: Generator-Classifiers For Supervised Language Modeling

Deconstructed Generation-Based Zero-Shot Model