AI Insight
This paper introduces Tunable MAGMAX, a framework for continual learning that allows customizable control over task-specific performance when merging multiple neural network models. The method uses preference vectors to determine how much each task's parameters contribute to the final merged model, enabling adaptation to different deployment environments with varying performance requirements. The framework can automatically generate appropriate preference vectors using small amounts of target environment data, eliminating manual tuning while maintaining or exceeding baseline performance across continual learning benchmarks.
Why it matters
This approach addresses a practical limitation in deploying machine learning models across diverse real-world environments where different tasks may have varying importance. By enabling automatic adaptation of merged models to specific deployment contexts without retraining, this framework could reduce computational costs and make continual learning systems more flexible for applications ranging from personalized AI assistants to domain-specific industrial systems.
Understand the Science
arXiv:2605.20803v2 Announce Type: replace
Abstract: Continual learning (CL) aims to train models sequentially on multiple tasks while mitigating catastrophic forgetting of previously learned knowledge. Recent advances in large pre-trained models (LPMs) and model merging techniques, such as MAGMAX, have demonstrated effective CL performance by combining task-specific parameters. However, existing methods primarily focus on average performance across all tasks and do not adequately address how to construct models accommodating different deployment environments or varying user preferences. This paper proposes a model merging framework, termed Tunable MAGMAX, which enables preference-aware control of task-specific performance in CL. Our method introduces a preference vector that controls the number of elements selected from each task vector during model merging, allowing us to adjust the merged model performance according to their deployment needs. We further propose a method for automatically constructing appropriate preference vectors by leveraging small amounts of target environment data and datasets from model training tasks, thereby eliminating the need for manual specification. The experimental result on CL benchmark tasks demonstrates that Tunable MAGMAX effectively controls task-wise performance and successfully adapts merged models to various target environments. The proposed Tunable MAGMAX achieves superior or comparable performance to baseline methods, making it a practical solution for deploying CL models to various environments where the preferences of each task performance differ.
Source: Tunable MAGMAX: Preference-Aware Model Merging for Continual Learning