Modular Delta Merging with Orthogonal Constraints: A Scalable Framework for Continual and Reversible Model Composition
This addresses the need for scalable, interference-free, and reversible model composition in real-world AI deployments, such as for compliance with GDPR, but it is incremental as it builds on existing merging and continual learning techniques.
The paper tackles the problem of continual model updates and composition in machine learning, which often leads to task interference and lack of reversibility, by proposing MDM-OC, a framework that encodes task-specific models as orthogonal deltas and merges them, resulting in improved accuracy, backward transfer, and unmerge fidelity on vision and NLP benchmarks.
In real-world machine learning deployments, models must be continually updated, composed, and when required, selectively undone. However, existing approaches to model merging and continual learning often suffer from task interference, catastrophic forgetting, or lack of reversibility. We propose Modular Delta Merging with Orthogonal Constraints (MDM-OC), a novel framework that enables scalable, interference-free, and reversible composition of fine-tuned models. Each task-specific model is encoded as a delta from a shared base and projected into an orthogonal subspace to eliminate conflict. These projected deltas are then merged via gradient-based optimization to form a unified model that retains performance across tasks. Our approach supports continual integration of new models, structured unmerging for compliance such as GDPR requirements, and model stability via elastic weight consolidation and synthetic replay. Extensive experiments on vision and natural language processing benchmarks demonstrate that MDM-OC outperforms prior baselines in accuracy, backward transfer, and unmerge fidelity, while remaining memory-efficient and computationally tractable. This framework offers a principled solution for modular and compliant AI system design.