Beyond Anti-Forgetting: Multimodal Continual Instruction Tuning with Positive Forward Transfer
read the original abstract
Multimodal Continual Instruction Tuning (MCIT) enables Multimodal Large Language Models (MLLMs) to meet continuously emerging requirements without expensive retraining. MCIT faces two major obstacles: catastrophic forgetting (where old knowledge is forgotten) and negative forward transfer (where the performance of future tasks is degraded). Although existing methods have greatly alleviated catastrophic forgetting, they still suffer from negative forward transfer. We discover a large discrepancy in different input embeddings by performing singular value decomposition (SVD) on input embeddings. This discrepancy results in the model learning irrelevant information for old and pre-trained tasks, leading to catastrophic forgetting and negative forward transfer. To address these issues, we propose Prompt Tuning with Positive Forward Transfer (Fwd-Prompt), a prompt-based method that projects the prompt gradient to the residual space to minimize interference between tasks and to the pre-trained subspace for reusing pre-trained knowledge. Our experiments demonstrate that Fwd-Prompt achieves state-of-the-art performance while updating fewer parameters and requiring no old samples. Our research illuminates the potential of continuously adapting MLLMs to new tasks under the instruction tuning paradigm and encourages future studies to explore MCIT.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
Modality-Inconsistent Continual Learning of Multimodal Large Language Models
The paper introduces the MICL scenario for MLLMs with modality and task shifts and proposes MoInCL using pseudo-target generation and instruction-based distillation, reporting gains over continual learning baselines o...
-
When the Same Musical Knowledge Forgets Differently: A Clean Probe of Pathway-Dependent Forgetting
Acquisition route affects forgetting rates in multimodal models, with text-pathway knowledge forgetting faster than audio-pathway knowledge in music understanding tasks.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.