FOREVER aligns replay intervals in LLM continual learning with a model-centric time based on optimizer update magnitudes and an Ebbinghaus-inspired forgetting curve to reduce catastrophic forgetting.
First five tasks correspond to the standard CL benchmark (Zhang et al., 2015)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
FOREVER: Forgetting Curve-Inspired Memory Replay for Language Model Continual Learning
FOREVER aligns replay intervals in LLM continual learning with a model-centric time based on optimizer update magnitudes and an Ebbinghaus-inspired forgetting curve to reduce catastrophic forgetting.