Stabilizing the Modality Gap: Null-Space Projected Continual Learning for Vision-Language Models
cs.LG, cs.CV
Submitted: 2025-07-26
Updated: 2026-10-13
Code: https://github.com/Ppp-Ttt/GNSP
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Efficient Lifelong Learning with A-GEM
- Microsoft COCO Captions: Data Collection and Evaluation Server
- Continual Multimodal Knowledge Graph Construction
- An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networks
- HiDe-LLaVA: Hierarchical Decoupling for Continual Instruction Tuning of Multimodal Large Language Model
- Distilling the Knowledge in a Neural Network
- Take Only What You Need: Rank Minimization as an Implicit Forgetting Regularizer in Continual Learning
- Fine-Grained Visual Classification of Aircraft
- Three scenarios for continual learning
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks