Back to home

Tag

catastrophic forgetting

Catastrophic forgetting is the tendency of a model to lose earlier skills or representations when it learns new tasks. It matters in continual learning, safe RL, long-sequence 4D reconstruction, and online adaptation, where retaining past behavior is as important as fitting new data.

3 articles

How to Prevent Catastrophic Forgetting in LLM Fine-Tuning
Research/Jun 6

How to Prevent Catastrophic Forgetting in LLM Fine-Tuning

Use Anchored Weight Decay to reduce prior-task drift during LLM fine-tuning.

Fixing LLM forgetting in ES fine-tuning
Research/Jun 5

Fixing LLM forgetting in ES fine-tuning

This paper shows LLM fine-tuning with evolution strategies can drift, and anchored weight decay can curb it.

Fast Spatial Memory for Long 4D Sequences
Research/Apr 9

Fast Spatial Memory for Long 4D Sequences

A new 4D reconstruction model uses elastic test-time training to reduce forgetting and memory bottlenecks in long-sequence spatial learning.