Adaptive Rank, Reduced Forgetting: Knowledge Retention in Continual Learning Vision-Language Models with Dynamic Rank-Selective LoRA
This addresses the problem of efficient continual learning without data storage for researchers and practitioners, though it is incremental as it builds on existing LoRA methods.
The paper tackles catastrophic forgetting in continual learning for vision-language models by proposing CoDyRA, which adaptively optimizes LoRA adapter ranks to balance plasticity and stability, achieving state-of-the-art results with improved knowledge retention.
Continual learning (CL) aims to accumulate knowledge from sequential tasks without catastrophic forgetting. Vision-language models such as CLIP, with strong generalization, are widely used for CL. Existing methods often adapt isolated PTM components, increasing inference complexity and limiting model improvement, or rely on replay, stored data, or assumptions, leading to high costs and limited applicability. To advance models as continual learners, we explore CL through natural and efficient PTM updates rather than complex task-specific additions. We study continual low-rank learning and analyze how LoRA ranks and placements affect learning and forgetting. A higher-rank LoRA improves task learning (plasticity) but increases forgetting, while a lower-rank LoRA enhances stability but limits adaptation. We observe a plasticity-stability balance tied to rank across parameters and tasks, with moderately small ranks maximizing CL benefits. Motivated by this, we propose Continual Dynamic Rank-Selective LoRA (CoDyRA), which continually updates PTMs with LoRA adapters of adaptively optimized ranks. The new-task objective drives learning, while sparsity-promoting regularization minimizes ranks to reduce interference and forgetting, achieving a balance tailored to each parameter and task. Although all parameters are updated, the minimized ranks keep the model close to its prior state while enabling effective new-task learning. CoDyRA performs efficient CL as a sequence of LoRA-based updates without storing past data or relying on assumptions, preserving the original model architecture and adding no inference overhead. Experiments show CoDyRA improves new representations while retaining old knowledge, achieving state-of-the-art results. Code is available at https://github.com/jeff024/codyra.