Prefer to Classify: Improving Text Classifiers via Auxiliary Preference Learning

Keyphrases

• Preference Learning 100%
• Text Classifier 100%
• Input Text 66%
• Model Accuracy 33%
• Input-output 33%
• Deep Neural Network 33%
• Cooperative Effects 33%
• Publicly Available 33%
• Low Pay 33%
• Pairwise Comparison 33%
• New Alternatives 33%
• Classification Task 33%
• Marginal Impact 33%
• Training Symbol 33%
• Human Cost 33%
• Additional Inputs 33%
• Crowd Worker 33%
• Auxiliary Data 33%
• Data Annotation 33%
• GPT-3 33%
• Large-scale Pre-trained Model 33%
• Multi-task Learning Framework 33%
• NLP Tasks 33%
• NLP Benchmark 33%

Computer Science

• Annotation 100%
• Preference Learning 100%
• Input/Output 33%
• Model Accuracy 33%
• Multitask Learning 33%
• Training Signal 33%
• Learning Framework 33%
• Generative Pre-Trained Transformer 3 33%
• Classification Task 33%
• Pairwise Comparison 33%
• Large Language Model 33%
• Deep Neural Network 33%