Self-Distillation Enables Continual Learning [pdf]

· HN · ArXiv ·

Researchers propose using self-distillation to enable continual learning in neural networks, mitigating catastrophic forgetting without external memory or regularization.

Categories: Research

Excerpt

HN · 78 points · 20 comments

Read at source: https://arxiv.org/abs/2601.19897

Discussions