ForkMerge: Mitigating Negative Transfer in Auxiliary-Task Learning
This addresses negative transfer for researchers and practitioners in multi-task learning, though it is incremental as it builds on prior optimization-based methods.
The paper tackles the problem of negative transfer in auxiliary-task learning, where learning multiple tasks simultaneously can reduce target task accuracy, and introduces ForkMerge, a method that periodically forks the model, searches task weights, and merges branches to filter harmful updates, achieving improved performance on benchmarks.
Auxiliary-Task Learning (ATL) aims to improve the performance of the target task by leveraging the knowledge obtained from related tasks. Occasionally, learning multiple tasks simultaneously results in lower accuracy than learning only the target task, which is known as negative transfer. This problem is often attributed to the gradient conflicts among tasks, and is frequently tackled by coordinating the task gradients in previous works. However, these optimization-based methods largely overlook the auxiliary-target generalization capability. To better understand the root cause of negative transfer, we experimentally investigate it from both optimization and generalization perspectives. Based on our findings, we introduce ForkMerge, a novel approach that periodically forks the model into multiple branches, automatically searches the varying task weights by minimizing target validation errors, and dynamically merges all branches to filter out detrimental task-parameter updates. On a series of auxiliary-task learning benchmarks, ForkMerge outperforms existing methods and effectively mitigates negative transfer.