English

I2I: Initializing Adapters with Improvised Knowledge

Computation and Language 2023-07-12 v2 Artificial Intelligence Computer Vision and Pattern Recognition Machine Learning

Abstract

Adapters present a promising solution to the catastrophic forgetting problem in continual learning. However, training independent Adapter modules for every new task misses an opportunity for cross-task knowledge transfer. We propose Improvise to Initialize (I2I), a continual learning algorithm that initializes Adapters for incoming tasks by distilling knowledge from previously-learned tasks' Adapters. We evaluate I2I on CLiMB, a multimodal continual learning benchmark, by conducting experiments on sequences of visual question answering tasks. Adapters trained with I2I consistently achieve better task accuracy than independently-trained Adapters, demonstrating that our algorithm facilitates knowledge transfer between task Adapters. I2I also results in better cross-task knowledge transfer than the state-of-the-art AdapterFusion without incurring the associated parametric cost.

Keywords

Cite

@article{arxiv.2304.02168,
  title  = {I2I: Initializing Adapters with Improvised Knowledge},
  author = {Tejas Srinivasan and Furong Jia and Mohammad Rostami and Jesse Thomason},
  journal= {arXiv preprint arXiv:2304.02168},
  year   = {2023}
}

Comments

Accepted at 2nd Conference on Lifelong Learning Agents (CoLLAs), 2023

R2 v1 2026-06-28T09:50:03.441Z