Do Pre-trained Models Benefit Equally in Continual Learning?

Keyphrases

• Training Model 100%
• Continual Learning 100%
• Model Benefits 100%
• Learning Algorithm 33%
• Learning Methods 8%
• Two-stage Training 8%
• Learning Objectives 8%
• Learning Scenario 8%
• Learning Communities 8%
• Knowledge Transfer 8%
• Underperformance 8%
• Art Performance 8%
• Dramatic Performance 8%
• Effective Baseline 8%
• Downstream Task 8%
• Performance Drop 8%
• Algorithm Learning 8%
• Training pipeline 8%

Computer Science

• Learning Algorithm 100%
• Regularization 50%
• Future Development 25%
• Art Performance 25%
• Learning Community 25%

Earth and Planetary Sciences

• State of the Art 100%
• Transferring 50%