Marius Dragoi
1 indexed paper
Recent (6 mo)
1With code
0Influential cites
0Benchmarked
0Publications per year
126
Top categories
ML×1
Frequent co-authors
Research Timeline
2026
TailLoR: Protecting Principal Components in Parameter-Efficient Continual Learning
TailLoR is a new parameter-efficient finetuning method that uses the singular bases of pre-trained weights to learn low-rank updates, specifically penalizing updates along dominant directions to improve continual learning.
Highlighted terms show continued research focus across papers