Can I read 2025 ICML BECAME Bayesian Continual L on EtoBox?

2025 ICML BECAME Bayesian Continual L by c3lttrong2a1btbinh is a document available to read on EtoBox.

What is 2025 ICML BECAME Bayesian Continual L about?

The document presents BECAME, a framework for Bayesian Continual Learning that enhances the stability-plasticity trade-off through adaptive model merging. It reformulates merging mechanisms using Bayesian principles and derives a closed-form solution for optimal merging coefficients, demonstrating superior performance over existing methods. Extensive experiments validate the effectiveness of BECAME in mitigating catastrophic forgetting while improving overall task performance.

Author

c3lttrong2a1btbinh

Language

EN